Methodology v1.0 · October 2026 research run, with the lab since 18 November 2026

The Agent Tool Benchmark

Every graded listing gets a score from 0 to 100 and a grade from AA to F. This page is the specification, with the checklist, the weights, and how our lab's measurements move a score.

Why another benchmark

The directories that exist rank agent tools by GitHub stars, package downloads, or how often a tool gets called through one particular proxy. Those numbers say what is popular. They don't say whether the tool answers at three in the morning, whether its descriptions make sense to a model, how many tokens it burns before the first useful call, whether an operator can hand it to an agent without handing over the keys, or whether an agent can pay for it without a human. Popularity and quality have already come apart. The most-used documentation server in the MCP world ships a 2,000-character tool description that a community grader marked as failing (we checked, and it is 2,006 characters).

We spent a decade grading crypto exchanges with a public method, and the thing we learned is that once the checklist is public, the people being graded start fixing the items on it. So the plan here is the same. Assess what can be read against a published checklist and show the working, probe what can be probed, and say plainly when something couldn't be verified.

An agent doesn't run on MCP servers alone. It runs on a model, inside a framework, reading data from providers and pages it scrapes, and sometimes paying for what it uses. So the benchmark covers all of those, on one scale, with each category read the way that makes sense for the kind of thing being scored.

What we benchmark

Five kinds of listing share the nine categories. The weights don't change between kinds. What each category measures does, and this table is the whole mapping. Open-weight models are the sixth kind, and since v0.6 (21 October 2026) they have a checklist of their own, with five categories that fit files you download and run, so their column here only points there. The reading goes with the kind, not with the directory category a listing sits in, so an open-weight speech model in Text-to-speech is graded by the open-weight checklist, the same as an open-weight chat model in Open-weight models.

CategoryMCP servers and APIsModel APIs, routers and model platformsOpen-weight modelsAgent frameworksAgent harnessesPayment protocols
ReliabilityStatus history, documented limits and overload handling, SLA (availability and error rate from probes once they run)Status history, overload errors and how they're signalled (probes once they run)Not used. Open-weight models have their own checklist (capability, reach, risk, agent interface, upkeep)Official package, CI, regressions and semver (task-suite runs once they run)Installs from an official package and runs headless without a harness errorReference implementations and public facilitators we can reach
Performancep50 and p95 of a representative callTime to first token and throughputNot used. Open-weight models have their own checklist (capability, reach, risk, agent interface, upkeep)Overhead per step on top of the model callTurns and wall time per task on top of the model callsTime from 402 to a settled, paid response
Schema & documentationTool descriptions and input schemasAPI reference, OpenAPI, llms.txt, structured outputsNot used. Open-weight models have their own checklist (capability, reach, risk, agent interface, upkeep)Docs and typed interfaces a model can followDocs, settings and permission rules an operator and a model can followSpec clarity, completeness and consistency with releases
Agent ergonomicsContext cost, pagination, recoverable errorsTool use, caching, context length, SDKsNot used. Open-weight models have their own checklist (capability, reach, risk, agent interface, upkeep)Code and defaults needed for a tool-calling agent with MCPWhat it takes to run headless in CI with one MCP server attachedWork for a seller to accept and a buyer to pay
Security & authScopes, read-only modes, confirmation on writesRetention, training on data, zero-retention optionsNot used. Open-weight models have their own checklist (capability, reach, risk, agent interface, upkeep)Telemetry defaults, approvals, guardrailsApproval modes, sandbox and network defaults, what leaves the machinePublished attacks and the controls that answer them
Payments & pricingMachine payment, published prices, card-free startPublished token prices, free tier, onboardingNot used. Open-weight models have their own checklist (capability, reach, risk, agent interface, upkeep)Licence and hosted-runtime pricingPrice of the harness and of the account it needs, and a free way to try itFees, rails and how much an agent can do without a person
Task successTask suite pass rate (data quality is half for data providers, once scored)Agent task suite with the model drivingNot used. Open-weight models have their own checklist (capability, reach, risk, agent interface, upkeep)The same task suite run through the frameworkThe repository task suite run through the harnessWhether an agent can complete a real purchase today
Maintenance & communityRelease cadence and responsivenessRetirement notice periods and model churnNot used. Open-weight models have their own checklist (capability, reach, risk, agent interface, upkeep)Release cadence and breaking changesRelease cadence, breaking changes and advisories handledSpec activity and governance
Transparency & trustEditorial plus the computed provenance scoreEditorial plus provenanceNot used. Open-weight models have their own checklist (capability, reach, risk, agent interface, upkeep)Editorial plus provenanceEditorial plus provenance, including telemetry disclosureEditorial plus provenance, including who governs the spec

MCP servers, HTTP APIs, model APIs, frameworks, agent harnesses, data providers and scraping tools are ranked together, because an agent builder does choose between them for the same budget. Open-weight models are ranked among themselves, by their own checklist. Payment protocols are graded on the same scale but not ranked against tools. You choose a protocol by who you're paying, not by its score, and putting a specification at number two above a search API told nobody anything useful. Retired listings keep their page and their grade and drop out of the ranking.

Categories and weights

Nine categories add up to 100 points. The weights follow what breaks agents in practice. A tool that is down is useless whatever its schema looks like, so reliability carries the most. A tool an agent can't pay for is a tool it can't use alone, so payments carry more than a payments category normally would. We're not certain the weights are right (they're our first guess, and the lab's measurements will tell us where they're wrong).

Seven of the nine are read from public evidence and make the total. Performance and Task success are measured by our lab, and since v1.0 they move a total up or down from there, as the lab section explains.

CategoryWeightWhat it measures
Reliability16%Does the tool answer, and does it keep answering the same way? In this run it's assessed from public evidence (90 days of status history, documented rate limits and overload handling, SLAs, general availability). Our own probes from three regions will replace that evidence as they accumulate.
  • 30-day availability of the endpoint (remote) or of a clean start plus tools/list (local)
  • Error rate on a fixed set of representative calls
  • Timeout rate and behaviour under back-pressure (429 with Retry-After, not silent hangs)
  • Consistency, meaning the same input gives the same shape of output across the window
  • Graceful degradation when upstream systems are down
Performance10%How long an agent waits, measured by our lab: the latency of each tool call it makes, from London. Measured, not read, so it moves a listing only once the lab has called its tools: above the neutral score it adds points, below it takes them away, and a listing the lab hasn't tested keeps the total it had.
  • p50 latency of the lab's tool calls (60%): full marks at 750 ms or less, none at 7.5 s or more, on a log scale between
  • p95 latency of the same calls (40%): full marks at 2 s or less, none at 20 s or more
  • Measured from London only, so the marks are set for a server on the far side of the world: an ocean of round trips doesn't cost a fast tool its score
  • A result below the neutral score counts once two runs on different days agree

Not tested by our lab yet, so it doesn't move this listing's total either way.

Schema & documentation13%Can a model understand what the tool does from the tool definition alone? We grade the text a model sees.
  • Tool descriptions state purpose, when to use, and when not to
  • Input schemas with types, enums, constraints and required fields; no free-form JSON blobs
  • Examples in descriptions or docs; documented error responses
  • Versioning of the tool surface and a changelog
  • Machine-readable docs (llms.txt, OpenAPI, registry server.json)
Agent ergonomics13%How much of an agent's context and how many round trips the tool consumes to get a job done.
  • Context cost, the tokens tools/list consumes before any work starts
  • Pagination, filtering and output-size controls
  • Actionable error messages an agent can recover from without a human
  • Idempotency and safe retries; read-only variants and tool annotations (readOnlyHint, destructiveHint)
  • Sensible defaults and few mandatory parameters; toolsets or meta-tools when the surface is large
Security & auth14%Can an operator give an agent this tool without giving it the keys to everything?
  • OAuth 2.1 with scoped, revocable, short-lived credentials; no secrets in URLs
  • Read-only and scoped modes; confirmation for destructive actions
  • Prompt-injection mitigations for tools that return untrusted content
  • Audit logging and per-call visibility for the operator
  • Handling of advisories (disclosed, patched, communicated)
Payments & pricing10%Can an agent start using the tool, and pay for it, without a human in the loop? A machine payment protocol on the tool's own endpoints is the largest single signal.
  • A machine payment protocol (x402, MPP or L402) on the tool's own endpoints (40 points)
  • Transparent, per-call pricing with units, published without a login (20 points)
  • A free tier or trial that doesn't require a card (20 points)
  • Autonomous onboarding, meaning an agent can obtain access without a human signup flow (20 points)
Task success10%Does every tool work? Our lab gives a model only what the server says about its tools and has it call each one; a tool works when it answers without an error. Full marks only when every tool works. For data providers, half of this category is the data-quality score, because a clean API over thin data still fails the task. Like Performance it moves a listing only once the lab has tested it, up above the neutral score and down below it.
  • The share of a server's tools that answered without an error, with any model the lab ran: 100 only when every tool works
  • A tool that changes something is called only in the dry run it declares (/agents/#dry-runs); while any tool couldn't be called, the score stays at 80 or below
  • A server that answers every call with 401 or 403 without a key is untested, not failed
  • For data providers, the data-quality score from /benchmark/#data-quality (half the category)
  • A result below the neutral score counts once two runs on different days agree

Not tested by our lab yet, so it doesn't move this listing's total either way. A data provider's data-quality score is published on its listing and becomes half of this category once it's tested.

Maintenance & community7%Is anyone home? Release cadence and responsiveness predict how a tool behaves after the protocol moves under it.
  • Release cadence in the last 90 days and time since the last release
  • Issue and pull-request responsiveness
  • Presence in the official MCP registry under a verified namespace
  • Package health (current SDK versions, no pinned-and-forgotten dependencies)
Transparency & trust7%Can an operator find out who runs the tool, what it does with data, and what changed? Half of this category is the provenance score, computed from checked facts. Legal entity, domain age, whether the endpoint sits on the vendor's own domain, terms, privacy policy, status page, changelog and security.txt, and since v0.7 the named people behind it and a support route, counted only when they raise the score.
  • Provenance score, computed from the checks on /benchmark/#provenance (half the category)
  • Source availability and licence clarity
  • Data handling and retention statements that agree with each other
  • Deprecation notices with dates
  • Telemetry disclosed and opt-out documented
Negative events−15 maxDeductions of up to 15 points for incidents in the last 12 months. Security incidents, breaking changes shipped without notice, silent deprecations, unresolved advisories, or misleading listings. Deductions are set by hand at each run, and a documented fix earns a smaller deduction or none; an automatic decay schedule is still open.
  • Confirmed data exfiltration path via the tool (up to -15)
  • Breaking tool rename or schema change without a deprecation window (-3 to -8)
  • Endpoint removed while still advertised in a registry or docs (-3 to -6)
  • Telemetry enabled without disclosure (-2 to -5)
  • Misleading listing or claim (-2 to -5)

The categories read from public evidence make the total, each weight divided by 80. Performance and Task success, measured by our lab, move it from there: each adds its weight × (score − 70) ÷ 100, up for a score above 70 and down below it, and a listing the lab hasn't tested isn't moved.

How this run was made

Until 1 October 2026 every score on this site was an invented sample, published to show the method. The October 2026 research run replaced them.

Between 30 September and 2 October 2026, research agents running on Anthropic's Claude models researched every graded listing from public evidence. Listings added or re-checked since 3 October 2026 were researched the same way, and each listing's assessment carries its own date. They read status pages and their incident history, rate-limit and error documentation, API references and OpenAPI files, MCP tool definitions in the source, pricing pages, terms and privacy policies, security and trust pages, changelogs, release tags and CI in cloned repositories, package registries and public advisories. They worked from the links on each listing and the pages those linked to, with a budget of about twelve page fetches per listing and no web search. Each category was scored against the checklist below. The points add up to the score, and the note beside each score on the listing says what earned them and where the agent departed from the checklist and why.

Five rules held for every listing.

  • Score from evidence found. Nothing gets points for what a vendor probably does.
  • Couldn't check isn't the same as absent. If a thing was looked for and isn't there (no status page, no SLA, pricing behind a login), it scores as absent. If a page couldn't be loaded, the agent could rely on the listing's own facts from their 30 September check and say so in the note. If those didn't cover it either, the item scores as absent, goes on the listing as an open question ("unchecked" and what), and the confidence drops. Our own fetch limits never read as a vendor's failing without saying so.
  • Vendor claims are claims. "Exa says its index covers 1.4 trillion URLs" is a fact about Exa's page, not our measurement.
  • No invented numbers, dates, incidents or quotes. If it wasn't found, it isn't true for this run.
  • Facts on a listing that turned out wrong (a renamed product, a moved endpoint, dropped x402 support) were corrected, and the correction is named in the listing's open questions.

Every graded listing carries three things beside its scores. A confidence, high, medium or low, which is how sure we are that the scores would survive the vendor reading the notes. The open questions, under "What we couldn't check". And the sources, at least four per listing, each with the date it was read. The JSON for a listing has all of it under anchor.assessment.

The conflict rule. The research agents and the review panel run on Claude, so Anthropic is a company we depend on. Anthropic's listings (the Claude API, Claude Code and the Claude Agent SDK) were graded by the same checklist as every other listing, with each judgement call written in the note, and each carries a disclosure. Since 15 October 2026, at our founder's request, the panel reviews them too, from the same dossier and by the same briefs, and the arbiter rules on those reviews. Each one carries a conflict label, and the listing warns that the reviewers run on the models of the company they're reviewing. Our own products are graded by the same readings as every listing, with two research agents and a third who reconciles them, LocalGhost since 3 October 2026 and letme since 20 October 2026 (below). From 2 to 20 October 2026 letme was graded by a stricter version of the same rule. A listing the founder built somewhere else, the CoinDesk Data API, is graded like the rest and carries a disclosure too.

The checklist

Each category has a checklist. The points are what an item is worth, and a listing's score in the category is the sum. Read each one the way that fits the kind of listing (the table above). These are the checklists the research agents used, written for a reader.

Reliability, from public evidence

Hosted APIs, MCP servers, models and platforms.

  • 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
  • 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
  • 15, rate limits documented with numbers.
  • 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
  • 10, an SLA published for any paid tier.
  • 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

  • 20, installs from an official package with supported runtimes stated.
  • 25, a public CI and test suite, passing on the default branch.
  • 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
  • 15, semver discipline and breaking changes called out in a changelog.
  • 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

Schema and documentation

APIs and MCP servers.

  • 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
  • 10, llms.txt or Markdown docs served for agents.
  • 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
  • 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
  • 0 to 15, examples and documented error responses.
  • 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

Agent ergonomics

  • 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
  • 20, pagination, filtering and output-size controls.
  • 20, actionable, documented error responses, codes and messages an agent can recover from.
  • 20, idempotency or safe retries, and for MCP the readOnlyHint and destructiveHint annotations.
  • 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

Security and auth

  • 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
  • 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
  • 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
  • 0 to 15, audit logs or per-call visibility for the operator.
  • 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

Payments and pricing

The published rubric, also on the x402 page.

  • 10, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 3 to 8 when it covers only some endpoints or only goes through a third party, and the note says which.
  • 30, per-call or per-unit pricing published without a login. 15 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
  • 30, a free tier or trial that doesn't need a card.
  • 30, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Until 13 November 2026 the first line was worth 40 and the other three 20 each. A tool with x402 and little else outranked tools that were better to use, so the protocol now counts for less than any of the basics. Every listing was re-scored by the same sum: its protocol points divided by four, and its other three lines times one and a half, with a line in each payments note saying so.

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 10 when x402, MPP or L402 runs on all their own endpoints, 8 when it runs on part of their own API, 6 when their merchants can accept one, 5 for running a facilitator, 4 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 30, 30 and 30 for the last three lines, and 0 to 10 for the first only if it ships a payment protocol.

Maintenance and community

  • 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
  • 20, at least three releases or dated changelog entries in the last 90 days.
  • 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
  • 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
  • 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

Transparency and trust, the editorial half

  • 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
  • 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
  • 0 to 20, a deprecation policy or notices with dates.
  • 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Open-weight models

An open-weight model is a set of files, not a service. Nobody holds a key, nothing is paid per call and nothing is down at three in the morning, so the nine categories above don't describe it well. Since methodology v0.6 (21 October 2026) open-weight models have a checklist of their own, with five categories that answer what an agent asks before it starts one. How good is it at the work, can I run it here, what could go wrong, how do I talk to it, and is anyone looking after it. Scores are 0 to 100 and the grade bands are the same as for every listing, but the two scales measure different things, so open-weight models are ranked among themselves and never against hosted tools. A model that also has a hosted API is two listings, the weights and the API, each graded by its own checklist and linked to the other (sameModel).

For an agent, these models are its cousins. One with spare hardware can start a cousin to take part of a job, and this checklist is there to say which cousin is good for what and whether it will run on the machine at hand. Every open-weight model by lab groups them by who trained them, with the family, older versions and the lab's own download link.

CategoryWeightWhat it asks
Capability35How well it does on the benchmarks agents depend on
Reach25Whether an agent can run it, on how much memory, with what, and on what terms
Risk20What could go wrong when you run it, and what the publisher has done about it
Agent interface10What a model and a developer need to talk to it
Documentation and upkeep10Whether it's described properly and someone is looking after it

The evidence is public. The model card, the files and their metadata wherever the publisher puts them, the licence, the chat template, the technical report, the runtimes' own code and documentation, who answered in the repository's discussions (we count replies and don't feed their text to a model), and benchmark results published by people other than the publisher. A gated model's card counts as what's public, and we don't accept its terms to see more.

Capability (35)

How well the model does on the benchmarks that matter to an agent. We don't average a publisher's own numbers into a grade. A result counts in full when we ran it on our reference machine, or when an independent party ran it and published the method (the Berkeley Function Calling Leaderboard, Artificial Analysis, Epoch AI's Benchmarking Hub, the SWE-bench and Terminal-Bench leaderboards, the Open ASR Leaderboard and others that run models themselves). A result the publisher submitted to such a leaderboard counts at half weight. A number that appears only on the publisher's card is shown on the listing as the publisher's claim and doesn't count.

Each result is scored as its share of the way from the benchmark's floor (chance for multiple choice, zero for a pass rate) to the best result any model, open or closed, has on the same leaderboard and setting on the day we read it. A model that matches the best model anywhere scores 100 on that benchmark. For an error rate such as word error rate, lower is better and the scale runs the other way, and where the leaderboard gives no floor the rate is read against the best model's, so twice the best model's error rate scores 50. Speed as a multiple of real time (RTFx) runs from about 1 to over 20,000, so it's read on a log scale from real time to the fastest model. On the MTEB leaderboard, where most models score between 0.5 and 0.75, the floor is the median of every model the same tab lists, and its results count at half weight because MTEB doesn't say who ran them and most are submitted with the model. An arena rating (Elo) isn't a ratio, so its floor is 400 points below the best model on that arena, the gap at which the better model wins about ten games in eleven. Where a model has results at several settings (reasoning effort, say), its best one counts, because that's the setting an agent would run.

The results group into areas, each weighted within the category.

  • Tool calling and agent tasks, 35 (BFCL, τ²-bench, Terminal-Bench and agent suites like them)
  • Coding, 20 (SWE-bench Verified, LiveCodeBench, Aider Polyglot)
  • Reasoning and knowledge, 20 (GPQA Diamond, MMLU-Pro, Humanity's Last Exam, composite indexes that publish their components)
  • Instruction following, 15 (IFEval, IFBench)
  • Long context, 10 (RULER, LongBench and long-context retrieval tests)

A speech model is read on its own areas, recognition accuracy (word error rate on the Open ASR Leaderboard's sets, 60) and speed (real-time factor, 40) for speech to text, and intelligibility and speaker similarity (Seed-TTS-eval, the Open TTS Leaderboard, 70) and speed (30) for text to speech. A publisher's own converted copy of the same weights (Qwen's -hf builds, say) counts as the same checkpoint. An embedding model is read on retrieval (60) and the other MTEB task types, classification, clustering, reranking and similarity among them (40), from the MTEB leaderboard. A vision model adds an image area (MMMU, ChartQA, DocVQA or OCRBench) at 20, and the others shrink to fit. A document parser, a model that turns page images into structured text, is read on the five things ParseBench measures, tables, charts, content faithfulness, semantic formatting and visual grounding, 20 each, from ParseBench's leaderboard or another independent document benchmark that publishes the same dimensions (since 22 October 2026). One leaderboard gives all five at once, so a parser's confidence counts independent sources, and a row the publisher submitted (or that the leaderboard copied from the publisher's card) counts at half weight with low confidence. LlamaIndex, which runs ParseBench, also sells LlamaParse, a parser the same benchmark ranks, and the listing says so.

Capability is the weighted mean of the areas with results, and needs results in at least two areas (one for a speech, embedding or document-parsing model). With less than that it's pending for that listing, its weight is shared across the other four, and the total is capped at 61, so a model nobody has benchmarked independently can't reach B, and every model with results ranked B or better stays above it. The listing says so. Until 21 October 2026 the cap was 69, which put models nobody had benchmarked above ones that had been.

Reach (25)

Whether an agent can run it, and where.

  • 0 to 30, memory. The smallest build from the publisher or a runtime's own library that keeps 4-bit precision or better, at its file size plus a fifth for the runtime and a working context. 30 for 16 GB or less, 25 for 24 GB, 20 for 32 GB, 15 for 64 GB, 10 for what fits our reference machine (the GPU share of 128 GB of unified memory), 5 for 256 GB, 0 above.
  • 0 to 25, runtimes. 5 for each of llama.cpp, Ollama, vLLM, SGLang, LM Studio and Hugging Face Transformers that runs the architecture in a released version, from the runtime's own code, library or documentation. A publisher's fork or an unreleased branch doesn't count. 25 at most.
  • 0 to 15, builds from the publisher. 15 for quantised builds at two precisions or more (FP8, 4-bit, GGUF), 10 for one, 0 when only other people's quantisations exist.
  • 0 to 15, terms of use. 15 for an OSI-approved licence with nothing added, 10 for a custom licence that lets anyone use it commercially for free below a threshold or under a use policy, 5 for research or non-commercial use only, 0 for no licence or unclear terms.
  • 0 to 10, download. 10 when the files download with no account and no terms to accept, 5 for a click-through that needs an account, 0 when a person approves each request.
  • 5, hosted too. The publisher's own API serves it, or at least two inference providers do, so an agent without the hardware can still call it.

When our reference-machine runs start, speed on that machine (tokens per second and time to first token at the build that fits) joins Reach.

Risk (20)

What could go wrong when you run it, and what the publisher has done about it. A higher score means less risk.

  • 0 to 25, file safety. 25 for weights in safetensors or GGUF only, 15 when pickle files (.bin, .pt, .pth) sit beside them, 0 for pickle only. 10 off when the documented way to load it or to build its prompts runs code from the repository (trust_remote_code, or the publisher's scripts in place of a template).
  • 0 to 15, provenance. 10 when the files come from the publisher's own organisation, the one its site or docs link to, and 5 for a checksum on every file.
  • 0 to 20, safety evaluation. 20 when results are published and someone other than the publisher ran them or can rerun them, 10 for the publisher's own results, 0 for none.
  • 0 to 15, guidance for agent use. 15 for documented prompt-injection guidance or an instruction hierarchy the model was trained to follow, 10 for a statement of what its trained refusals cover and the safeguards to add around it, 5 for general cautions, 0 for none.
  • 0 to 15, licence risk. 15 to start, 5 off for each of these, a use policy or field-of-use limits beyond the law, terms the licensor can change or revoke, a revenue or user threshold, and a licence that changed without a dated note or whose link is broken. Not below 0.
  • 0 to 10, a security programme. 5 for a way to report a flaw (security.txt, a disclosure policy or a security contact), 5 for advisories handled in public.

When our reference-machine runs start, measured robustness (prompt injection with AgentDojo, over-refusal with XSTest) joins Risk.

Agent interface (10)

What a model and a developer need to talk to it. Speech, image and other models with no chat interface read the tool-calling and structured-output lines for what the model takes and returns (voices, timestamps, streaming, typed outputs), and the note says how.

  • 0 to 20, the chat template. 20 for a template in the repository (Jinja, in the tokenizer config or its own file), 10 when the prompt format is only given as code you run.
  • 0 to 25, the tool-calling format. 25 when the publisher documents how tools are declared and how calls come back, with the parser to use in a named runtime, 10 when only the template or a runtime's own recipe says it.
  • 0 to 20, tool calling the model was trained for. 20 with several calls in one turn, 10 for one call at a time.
  • 0 to 15, structured output. 15 when JSON or JSON Schema output is documented for the model with how to ask for it, 5 when the card claims it without saying how.
  • 10, reasoning control. Thinking that can be turned off or budgeted per request, with where the reasoning text comes back.
  • 0 to 10, context. 10 for 128K tokens or more trained, 5 for 32K to 127K. Context reached only by scaling the position encoding counts one band down.

Documentation and upkeep (10)

  • 0 to 20, the model card. Intended use, limitations, architecture, parameters, context length, languages and knowledge cut-off. 20 for all seven, 3 off for each that's missing.
  • 0 to 20, training data. 20 for the data published or listed with sources and sizes, 15 for a data card that names the sources or the mix, the cut-off and the filtering, 10 for a general description with a cut-off, 5 without one, 0 for none.
  • 0 to 20, the last change to the files, the template or the reference code. 20 within 30 days, 15 within 90, 10 within 180, 5 within a year, 0 older.
  • 0 to 20, responsiveness. The publisher's replies to the 40 most recent questions and reports in the repository's discussions or issues. 20 for most, 10 for some, 5 for a few, 0 for none.
  • 0 to 10, version policy. 10 for earlier versions kept in their own repositories with a dated note when one is replaced, 5 for earlier versions kept without notes or a first release, 0 when files were replaced in place without a note.
  • 0 to 10, file health. 10 when config, tokenizer, template and card agree and no report of a broken file waits without a reply, 5 when one does, 0 when the files contradict the card.

Deductions, confidence and our own runs

The negative events above apply as for every listing, 0 to −15 (a revoked licence, a tampered or malicious file in the publisher's own repository, a safety incident the publisher confirmed). Confidence is high when Capability rests on independent results in three or more areas, medium for two, low otherwise.

Our reference machine is an AMD Ryzen AI Max+ 395 with 128 GB of unified memory, running Linux, llama.cpp and vLLM. When its runs start, our own results count in full in Capability, speed joins Reach and measured robustness joins Risk, every listing's history records the change, and a model too big for the machine keeps the independent results it has. The plan is at /benchmark/open-weights-runs/.

Performance and Task success, from our lab

Until v1.0 these two were pending, because neither can be read from a vendor's pages. Since 18 November 2026 our lab measures them. A model is given only what a server says about its tools and writes programs to call each one, and the lab's gateway counts and times every call, from London. The harness is open source, and what it found is on each listing it has tested, under Lab, and at /api/v1/lab.json.

The lab can only test what it can call: today that's hosted MCP servers that list their tools without a key, and ones whose vendor gave us a key. So most listings haven't been tested, and the rule is built for that. A tool that tested well should rank above the same tool untested, and an untested one above one that tested badly. Untested listings shouldn't be marked down for it, and they shouldn't be rewarded either.

That rules out the usual weighted mean. If a missing category's weight were shared across the rest, a tool with strong scores elsewhere would drop for testing merely well, and a weak one would rise for testing merely average. So the seven categories from public evidence make the total as before, and the two from the lab move it from there, around a neutral score of 70.

total = Σ (category score × weight) ÷ 80, over the seven read from public evidence
      + Σ weight × (lab score − 70) ÷ 100, over Performance and Task success, where tested
      then the negative events, 0 to −15

With a weight of 10 each, a listing gains at most 3 points from each (a score of 100) and loses at most 7 (a score of 0). The downside is bigger because a tool that's broken when an agent calls it costs that agent more than a fast one saves it. A listing the lab hasn't tested keeps the total it had.

Performance

The latency of the lab's tool calls: p50 counts 60% and p95 40%. Full marks for a p50 of 750 ms or less and a p95 of 2 s or less, none for 7.5 s and 20 s or more, on a log scale between. We measure from London only, and we're based in the UK, so the marks are set for a server on the far side of the world. Several ocean round trips shouldn't cost a fast tool its score.

Task success

Does every tool work? A tool works when it answered without an error, with any model the lab ran, so a weaker model's mistakes don't count against the server. The score is the share of a server's tools that work, and 100 only when all of them do: fourteen of fifteen isn't full marks.

A tool that creates, sends or deletes is called only in the dry run it declares ("_meta": {"anchorterminal.com/dryRun": true} and a dryRun argument, on /agents/#dry-runs), because calling it for real on a production server would publish, send or delete something. While any of a server's tools couldn't be called, its Task success stays at 80 or below. A server that answers every call with 401 or 403 is untested, not failed. For data providers, half of Task success is the data-quality score already published on their listings.

When a result counts

A result above 70 counts at once. One below 70 counts once two runs on different days agree, so a server caught mid-deploy isn't marked down for it. Every move is in the listing's history with the cause "New evidence" and the run it came from.

Days without breaking an agent

Every listing we probe shows how many days it has gone without breaking an agent. An outage our probes saw (three failed probes in a row, five minutes apart) or a dated breaking change or shutdown on the listing sets it back to 0. A price change, a rename or a notice doesn't. The count starts from the day we started probing the endpoint, so a breaking change from before that doesn't stand for a run we didn't watch. The longest run counts only what we still hold, which is 90 days of outages. It's in each listing's JSON as streak, and it doesn't change the score.

Our own products

letme (letme.dev) is Anchor Terminal's own product, and LocalGhost is built by its founder. Until 2 October 2026 we listed them and didn't grade them. On 2 October both were graded by the same checklist as every listing, with four differences (the third ended on 5 October 2026), because a grade we give ourselves has to hold up for a reader who assumes we were kind to ourselves. LocalGhost left that rule on 3 October 2026 and letme on 20 October 2026 (below).

  • Two research agents graded each one independently, and every judgement call was read strictly. Only what was live and public that day counted. Anything specified, planned or "coming" earned nothing, and so did our own description of how something works without the code, the live answers or a dated document behind it.
  • A third agent, which graded neither, reconciled the two item by item on the checklist. The lower award stood unless it rested on an error the auditor checked for itself, and a deduction counted only if the auditor could still see the problem that day.
  • Until 5 October 2026 the review panel didn't review them. Since then, at the founder's request, it reviews them like any listing (below).
  • letme never picks them and never answers with them, so it can't favour the company that runs it. Their records say own: true. They're in rankings, categories and comparisons like any listing.

Since 3 October, at the founder's request, LocalGhost is graded the same way as every listing instead, with the same readings, neither stricter nor looser, the way a competitor of ours is. Two research agents grade it independently and a third reconciles them, checking the evidence itself wherever they disagree, with no award standing by default. letme still never picks it, because that protects against us favouring ourselves rather than marking us down.

letme moved to the same rule at its next check, on 20 October 2026, also at the founder's request. Two research agents graded it from what was live and public that day, starting from a copy of its listing with the earlier grade taken out, and a third reconciled them item by item the same way. Anything planned or "coming" earned nothing, and our own description of how something works counted only with the live answers or a dated document behind it, as for any listing. It was graded again the same way on 23 October 2026, at the founder's request, from what was live that day. letme still never picks itself. The listing says what moved and why.

Mumble (tinyspace.ai/mumble), made by tiny space, the company of Quynh Tran-Thanh, a founder of Anchor Terminal, was listed on 14 October 2026 and graded the way LocalGhost is, with the same readings as every listing, two research agents and a third who reconciles them, and own: true, so letme never picks it.

Since 5 October 2026, at the founder's request, the review panel and the audience reviewers review letme and LocalGhost like any listing. Each reviewer is told the listing is our own product and that its review must be neither kinder nor harsher than for any other, and it writes from the same dossier and the same brief as every desk review. The reviews are shown on the listing like any other, with the same labels. letme's were written again on 20 October 2026 from the new dossier, by all eight panel lenses and the six audience reviewers, and the arbiter ruled on them.

Each carries a disclosure that says how it was graded, the reason and sources for every score are on the listing like any other, and anyone can dispute an item the same way. Every grading and reconciliation is kept with the research run.

Listed, not graded

A listing can be listed and not graded (graded: false), with no score, grade, rank, reviews or letme pick, and a disclosure that says why. None is today.

Competitors of our own products

When a listing competes directly with one of our own products, a reader could suspect the founder's side marked it down, or that we went easy on it to look fair. So it's graded by the same checklist as every listing, neither stricter nor looser, and the grading is set up so that neither can creep in.

  • Two research agents grade it independently. Where a checklist line could be read two ways, they take the reading already used for the same line on comparable listings. They don't look at our own product's listing and don't compare the two.
  • A third agent, which graded neither, reconciles them item by item. Where the graders agree, their award stands unless the auditor finds it rests on an error. Where they disagree, neither award stands by default. The auditor checks the evidence itself and awards what it supports.
  • What we couldn't read counts as absent, as for every listing, and the note says when a score is low because we couldn't read something rather than because it's missing. When a vendor's site refuses our reader, text supplied to us from it counts, and the listing says who supplied it.
  • The review panel reviews it like any listing, and letme can pick it.
  • A disclosure at the foot of its page, and of any comparison it's in, says which of our products it competes with and how it was graded. Its record says competesWith and names that product.

35 listings are graded this way today. 33 compete with LocalGhost: Underdog was the first, on 3 October 2026, Khoj, Open WebUI and Screenpipe followed the same day, the three that LocalGhost's own about page names as its competitors, and Ghost Core and the personal assistants listed from 12 October followed. Two compete with letme. Nevermined, whose Router added on 15 October 2026 picks a service for a task and pays for it, which is letme's job, and Monid, listed on 23 October 2026, which finds and runs paid endpoints for an agent from one balance. A listing is graded this way when our product names it as a competitor, or when it does the same job for the same people, as Underdog does. Both gradings and the reconciliation are kept with the research run for each.

Negative events

Scores describe the steady state. Negative events describe what happened. Up to 15 points come off for incidents in the last twelve months that can be sourced. A security incident or an exfiltration path, up to 15. A breaking change shipped without notice, 3 to 8. An endpoint removed while still advertised, 3 to 6. Telemetry nobody disclosed, 2 to 5. A misleading listing or claim, 2 to 5. Deductions are set by hand at each run, and a vendor that documents the fix and ships mitigations gets a smaller deduction or none. An automatic decay schedule is still open (below). A vendor that ships a breaking rename in a minor release and says nothing carries the full term. Every deduction is written on the tool's page with a date and a source.

Warnings

Some listings are real and usable and still look wrong. Since 13 November 2026 those carry a warning, shown above the verdict on the page, in its Markdown twins and in the API, and each warning takes its points off the total on top of any negative events, up to 30 points in all.

  • Nobody found behind it, 15. A hosted service that takes data or money, where no person and no company could be found on any of its own pages: no named founder or team, no legal entity in its terms, privacy policy or footer.
  • Operator not identified, no points off. A hosted service that takes data or money and names a company on its own pages, where the company still can't be identified: no place of registration, registration number or address, and no entry found in a company register.
  • Breaks the rules of the platforms it acts on, 15. Sold to do what the platforms it acts on forbid: making, warming or selling accounts on a social network whose terms ban it, fake engagement, ban evasion.

Since 16 November 2026 a listing with either of the first two can't be graded above B, so it isn't called agent-ready: its total stops at 69.9 after every deduction. A service whose operator nobody can identify has nobody accountable for what it does with an agent's data or money, however good its documentation.

We list these rather than hide them. Hiding one leaves people to meet it somewhere else and assume it's fine; listing it puts the reason in front of everyone who looks it up. A warning comes off when the finding no longer holds, a company behind the service that can be looked up, say, and the score history records the move. A warning is set by the research, with its sources, never worked out from missing fields alone: a listing whose pages we couldn't read gets no warning for what we couldn't see.

Grades

GradeScoreMeaningTools in this run
AA85+Exceptional. Agent-ready, with agent-native payments or equivalent autonomy.8
A78+Excellent. Agent-ready; minor gaps.53
BB70+Good. Agent-ready with documented caveats.254
B62+Usable. Needs operator supervision or workarounds.410
C54+Mixed. Material gaps in one or more categories.426
D46+Poor. Not recommended for autonomous use.251
E38+Very poor.147
Fbelow 38Failing or unverifiable.159

Agent-ready means BB or better, which is our shorthand for "you can give this to an autonomous agent as long as you read the caveats on its page". Below BB expect to supervise it, patch around it, or replace it. F is for tools that fail the checklist across the board or can't be verified at all.

Checked dates and score changes

Terms and prices change all the time, and an old score looks the same as a fresh one. So every score on a listing says the day its evidence was last checked, and a listing whose oldest check is more than 90 days old is marked as due a re-check. A re-check that covers only some categories moves only their dates.

Scores move. A vendor rewrites its terms, raises a price, has an outage or ships a release, a page we couldn't read becomes readable, our checklist changes, or we got something wrong. Each listing keeps its score history, and every number that moves carries a cause and a sentence saying why. The causes are terms, pricing, reliability, security, product (the vendor changed the product), company (who's behind it), new evidence (we read something we hadn't), correction (our mistake, or a reading brought into line with other listings), checklist (our method changed), negative event and time (a dated check, like the domain's age, crossed a line). A score can't change without an entry: the build compares every listing with the last entry in its history and fails if a number moved and nobody wrote down why. Only time is worked out on its own, from the date.

Each listing shows its history under its score, and score changes has every move across listings, newest first, with an RSS feed. The history starts with the first graded run (2 October 2026). The moves between then and 21 October were rebuilt from the published data at each change, with their reasons taken from the re-check notes written at the time.

Who's behind it

Credibility shouldn't be a feeling. For every listing we record the legal entity named in its terms, its registrable domain and the date the registry says it was first registered, whether the hosted endpoint sits on a domain the vendor controls, and whether there are terms, a privacy policy, a status page, a changelog and a valid security.txt. Since v0.7 (27 October 2026) two more checks come from each listing's Humans behind the project section, whether the people who run it are named with their role and whether there's a published route to get support. They count only when they raise the score, because a listing nobody has researched yet shouldn't lose points for it. Each check is worth fixed points and the total is the provenance score out of 100. It's half of Transparency and trust. The other half stays editorial, because licence clarity, retention statements that agree with each other and honest telemetry disclosure need reading, not counting.

CheckPointsHow it's scored
Legal entity named20The terms, imprint or licence name the company or body that stands behind the service.
Domain age15Years since the registrable domain was first registered, from the registry's own RDAP server. 10 years or more scores 15, 5 scores 11, 2 scores 7, 1 scores 3. Government domains score in full. Registration can predate the current owner, which is why this never stands alone.
Endpoint on the vendor's domain15The hosted endpoint sits on a domain the vendor controls, not a shared host or a lookalike. Not scored for libraries, specifications and local servers.
Terms of service10Published and reachable without a login scores 4. Once we've read it, the other 6 are shared between the seven things a reader expects it to state (its date, governing law, a liability limit, how it ends, how changes are announced, what users may not do, a service level), and each clause that restricts automated access, restricts benchmarking, lets the terms or service change without notice, or lets customer content train models with no opt-out takes 2. A document our reader was refused scores 7. For software you run yourself with nothing hosted, an open-source licence counts in full.
Privacy policy10Published and reachable without a login scores 4. Once we've read it, the other 6 are shared between the eight things a reader expects it to state (its date, what is collected, how long it's kept, who else receives it, whether it's sold or used for advertising, people's rights, a contact, where it's transferred), and a clause that lets customer content train models with no opt-out takes 2. A document our reader was refused scores 7. Not scored for software you run yourself with nothing hosted.
Status page10A public status page with incident history.
Changelog10Dated release notes or a changelog an agent can read.
security.txt10A valid RFC 9116 security.txt on the vendor's domain. An expired one scores half.
People named10The founders or the people who run it, named with their role from a first-hand page (the company's own site, Y Combinator, a registry), as in the listing's Humans behind the project. Named people score 7, and 10 when at least one of them has a public profile (LinkedIn, X or GitHub) the research tied to them. Not scored until the research has named someone, and like the support route it only counts when it raises the score.
Support route5Somewhere to get help that the company publishes, a support address, a help centre or a community such as Discord, scores 5. Only a general contact address or page scores 2. Not scored until the research has looked, and only counted when it raises the score.

Terms and privacy, as read

Since 8 October 2026 we read the terms of service and the privacy policy each listing links to, and the two checks score what the documents state, where before they scored only that a document was published. A fixed set of rules reads the text. Seven look for what a reader expects in terms of service (the date it was last updated, the governing law, a limit on liability, how the agreement ends, how changes are announced, what users may not do, a service level). Eight look for what a reader expects in a privacy policy (its date, what is collected, how long it is kept, who else receives it, whether it is sold or used for advertising, people's rights, a contact, where it is transferred). A few more look for clauses worth knowing before an agent relies on the service. Each answer is the vendor's own sentence, quoted on the listing.

A published document scores 4 of its 10. Once read, the other 6 are shared between the expected items. Four kinds of clause take 2 each, because they bear on an agent directly. They are a restriction on automated access, a restriction on benchmarking or competitive use, a right to change the terms or the service without notice, and a licence to train models on customer content where no opt-out is found. Arbitration clauses, an account that can be closed without notice and a document older than three years are shown and cost nothing. A document our reader was refused (robots.txt, a browser check, a page drawn by script) scores 7, so a vendor is neither rewarded nor penalised for a page we couldn't read, and the listing says why.

A rule can miss a clause or misread a sentence. Every clause that costs points was checked against the text by a second reader, and a hit that turned out to be about something else is set aside in the data file with the reason, where anyone can see it. None of this is legal advice. The document is what binds, and the listing links to it.

Domain age is the weakest of these and we know it. A registration date is when the name was first taken, not when the current owner bought it, so slack.com (1992) and notion.com (1997) look older than the companies are. Every listing where that's true says so under the checks. That's why age is 15 points and never decides a grade on its own. A brand-new domain with a named company, published terms and a working status page still scores well, and a very old domain with nobody's name on it doesn't.

Registry dates come from each registry's own RDAP server. Where the registry doesn't publish one (.gov.uk, some brand top-level domains) we say so on the listing and score government domains in full.

Data quality

A clean API over thin data still fails the task. For data providers, half of Task success is a data-quality score computed from what the data covers, how fresh and how deep it is, whether the method and sources are published, what the licence lets you do, and whether the publisher has official standing. Coverage is the one judgement call in it, rated 1 to 5 with the evidence written next to the number. It's computed and published on every data provider's listing, and it counts in the total once the lab has tested the listing, as half of its Task success. Each listing also says what its terms allow for caching and redistribution, because an agent that stores a price it wasn't allowed to store is a problem for its operator.

CheckPointsHow it's scored
Coverage25Breadth and depth of what the data covers against the obvious alternatives, rated 1 to 5 by the panel and published with the evidence.
Freshness15Real time 15, minutes 13, hourly 10, daily 7, static 3.
History1520 years or more 15, 10 years 12, 5 years 9, 1 year 5, less 2.
Methodology published15How the data is collected, cleaned or calculated, in public.
Sources disclosed10Where the data comes from, named.
Licence10Open licence 10, clear proprietary terms 7, unclear 2.
Official standing10Statutory register 10, regulated administrator 9, none 0.

How each category is measured

In this run, from public evidence

Reliability, schema and documentation, agent ergonomics, security, payments, maintenance and the editorial half of transparency were scored against the checklists above by research agents reading what a model reads and what an operator would check. The tool definitions, the docs, the auth flow, the status history, the pricing, the terms and the repository. Each score's note is on the listing, with the sources. Vendors can dispute any item with evidence. Disputes are answered by email within ten working days, and when the evidence holds, the score changes with a dated note in the listing's score history.

Machine payment in this run was read from each vendor's documentation, its source and its listing in the x402 Bazaar where it had one.

In the lab, by calling the tools

Performance and Task success come from the lab, as above. It runs on one machine in London, with open-weight models, and tests each server again every seven days and with each new model.

Still to come

Remote tools will be probed every five minutes from London. A probe opens a session, lists tools, and runs one small fixed call that doesn't change state. Local tools will be started in a clean container every fifteen minutes, listed, and called once. We'll record availability, latency percentiles, error classes, timeouts, and whether the response shape matched the previous run. The token count of tools/list gets measured with a reference tokeniser in the default configuration and again with every optional toolset switched on, and both numbers are published, because the gap between them is the first thing to configure.

Machine payment will be detected, not read. For x402 we call the endpoint without paying and look for a 402 carrying a PAYMENT-REQUIRED header (v2) or a JSON body with x402Version: 1 (legacy) that names a scheme, a CAIP-2 network, an amount, an asset and a payee. MPP (WWW-Authenticate: Payment) and L402 (WWW-Authenticate: L402) challenges get detected the same way. For tools that settle, one real micropayment a week from a canary wallet, with the transaction hash published.

Each category will have a suite of representative tasks (find a fact on the web, open a pull request, run a read-only query, pull a table out of a page). Once a month the suite runs through every tool in the category with a fixed reference model and a fixed harness, and we record pass@1, pass@4, turns, tokens, and whether the tool's own error messages were enough to recover from a first failure. The suites will be published so vendors can run them.

Maintenance and community, by public signals

Release cadence, time since the last release, responsiveness on issues and pull requests, presence in the official MCP registry under a verified namespace, and dependency health, all read from public repositories and package registries on the run date. For model APIs this is mostly about retirements. How much notice, how often, and whether a retired id fails loudly. Every dated change we find goes on the listing and on Sunsets, with a calendar feed at /sunsets.ics.

Leaderboard

The full ranking with every assessed category score. Sort and filter in the directory, or fetch it as JSON at /api/v1/rankings.json.

#ToolCategoryGrade Score ReliPerfScheAgenSecuPaymTaskMainTranNeg
1AI SDKFrameworksAA93.693–100959090–97890
2Vertex AI (Gemini Enterprise Agent Platform)ModelsAA91.299–94979760–91900
3OpenAI Agents SDKFrameworksAA9085–95978090–100890
4InfisicalSecretsAA86.7100–87919153–90840
5OpenAI APIModelsAA85.875–1009810045–87900
6GPT-6.2 AstraModelsAA85.875–1009810045–87900
7Pydantic AIFrameworksAA85.683–100959590–9087-6
8Stripe API + MCPPlatformsAA8578–90919766–82870
9Composio (API + MCP)Tool accessA84.470–89908490–93830
10MongoDB MCP ServerDatabasesA83.988–81758190–94840
11Backblaze B2StorageA83.890–89859060–80830
12Strands AgentsFrameworksA83.787–95878990–8772-4
13Amazon BedrockModelsA83.580–94929845–81850
14Stripe IdentityVerificationA83.277–94938660–85850
15Microsoft Agent FrameworkFrameworksA83.178–95879390–9185-5
16Google Calendar APISchedulingA8290–88917753–87820
17Amazon Bedrock AgentCore MemoryMemoryA81.688–92868745–83790
18Nylas Email APIMailboxesA81.677–93818960–88820
19Browser UseBrowserA81.587–86786998–82680
20MCP Toolbox for DatabasesDatabasesA81.480–85827590–95640
specMachine Payments Protocol (MPP)Pay per callA81.385–89917991–9750-3
21Resend API + MCPEmailA81.290–95676975–93830
22Azure OpenAI in Foundry ModelsModelsA81.280–87989530–86830
23Gmail APIMailboxesA80.990–87828153–85820
24Descope Agentic Identity HubAgent authA80.6100–78808660–74680
25AWS Secrets ManagerSecretsA80.377–96909145–70800
26Bird API + MCPMessagingA80.370–92958253–83880
27Basic MemoryMemoryA80.183–89856175–91820
28Twilio API + MCPMessagingA79.975–92808560–85810
29Twilio Programmable Voice API + MCPCallingA79.975–92808560–85810
30Mapbox APIs + MCPMapsA79.782–85887160–90840
31Tavily API + MCPSearchA79.590–90815294–85620
32AgentOSFrameworksA79.577–92748190–8770-2
33KnockNotificationsA79.483–93807175–86610
34Amazon Bedrock GuardrailsGuardrailsA79.375–92939645–65700
35Pinecone API + MCPRetrievalA79.380–94837960–9084-2
36Cloudflare Workers OAuth ProviderAgent authA79.374–86748075–87860
37Amazon S3StorageA79.280–92838845–73840
38Claude APIModelsA79.260–90959245–91850
specx402Pay per callA79.287–86846396–9369-3
39Claude Fable 5.2ModelsA79.260–90959245–91850
40HookdeckWebhooksA79.185–90816775–74770
41Apify MCP ServerScrapersA7975–92926299–9586-5
42BlockRun.AIModelsA78.965–908861100–85730
43Cloudflare R2StorageA78.882–92908530–87750
44Context.devScrapersA78.867–96908375–8578-3
45ElevenLabs Voice Cloning and Voice Design APICloningA78.773–91888445–85830
46Google Sheets APISpreadsheetsA78.690–83748053–85800
47Google Cloud StorageStorageA78.680–81858545–88850
48OpenAI Decisions APIDecisionsA78.570–91878745–79890
49ZenRowsScrapersA78.487–937950100–9079-3
50Amazon Location ServiceMapsA78.495–87887645–60770
51Customer.ioLifecycle marketingA78.475–86837968–80760
52MeilisearchRetrievalA78.390–93876760–9376-3
53Amazon Bedrock AgentCore IdentityAgent authA78.285–88768445–85780
54Firecrawl MCPScrapersA78.178–91927075–8775-3
55Speechify API Voice CloningCloningA78.190–94978730–7575-3
56Deepgram Text-to-Speech (Aura-2, Flux TTS)TTSA78.180–95827260–73780
57Amazon Bedrock model customisationFine-tuningA7895–87778745–45860
58Nylas Calendar and Scheduler APISchedulingA7872–91718560–88820
59MastraFrameworksA7887–98926790–9672-8
60AhaSend API + MCPEmailBB77.995–86807860–8474-3
61Google Maps Platform + Grounding Lite MCPMapsBB77.990–77857360–80720
62Google Drive API + MCPStorageBB77.870–89878353–85750
63DiditVerificationBB77.890–88697690–8264-3
64ElevenLabs Text to Speech API + MCPTTSBB77.880–94926560–77680
65Robot FrameworkDesktopBB77.795–86785090–60780
66TempoPlatformsBB77.666–90946083–88680
67Shopify API + MCPCommerceBB77.478–92837860–8589-3
68Docker AgentFrameworksBB77.488–90817090–9681-7
69OpenAI CodexHarnessesBB77.355–90808290–8785-2
70Sequential Thinking (MCP reference server)ReasoningBB77.380–80737090–72760
71Claude Mythos 5.2ModelsBB77.360–90959230–91850
72Amazon PollyTTSBB77.280–90828345–70800
73You.com APIsSearchBB77.280–92835997–8567-3
74Amazon Nova Multimodal EmbeddingsEmbeddingsBB77.195–76789145–50820
75Azure Blob StorageStorageBB77.188–85848130–80810
76Google Ads APIAdvertisingBB77.182–84876160–87820
77Microsoft Execution ContainersSandboxesBB77.181–81746990–9284-3
78Web Search Tool on Amazon Bedrock AgentCoreSearchBB77.190–78758545–80770
79Gemini Exchange APIExchangesBB77.173–88797860–88750
80GroqCloudModelsBB7790–64808260–72850
81OrbPlatformsBB7787–90847245–84660
82ScrapflyScrapersBB7785–95807560–8475-3
83Google Cloud Secret ManagerSecretsBB76.987–83828030–87840
84Amazon TranslateTranslationBB76.990–90807845–65720
85Deepgram Speech-to-Text (Nova-3, Flux)STTBB76.980–95757260–73780
86OpenAI Speech to TextSTTBB76.865–89789745–71900
87DatoCMSCMSBB76.780–85787753–81800
88TemporalHuman approvalBB76.680–91858830–9091-3
89Typesense API + MCPRetrievalBB76.683–86936560–8279-2
90Vercel SandboxSandboxesBB76.675–88738675–8278-3
91Amazon EventBridgeWebhooksBB76.578–82738953–65890
92Cloudflare AI SearchRetrievalBB76.570–89886868–78740
93LiveKit AgentsVoice agentsBB76.567–81806290–87820
94Saleor API + MCPCommerceBB76.575–89867168–8576-2
95HubSpot API + MCPCRMBB76.475–92728245–84850
96Azure AI Speech speech-to-textSTTBB76.390–80758530–80860
97AWS End User MessagingMessagingBB76.373–88718845–85850
98Chrome DevTools MCPBrowserBB76.283–87736390–9382-4
99Supabase API + MCPDatabasesBB76.260–89788453–90900
100Auth0 for AI Agents (Token Vault)Agent authBB76.280–85718845–74840
101HeyGen VoiceTTSBB76.175–97886960–65660
102Amazon TextractDocumentsBB7696–82707453–58850
103Turnkey Agentic WalletsWalletsBB7683–79778568–8373-3
104Arize PhoenixEvalsBB75.873–88905890–8475-3
105Parallel Search and Task APIsSearchBB75.775–94874581–87650
106Microsoft Entra Agent IDAgent authBB75.791–87718330–80750
107Akeyless (SecretlessAI and MCP server)SecretsBB75.690–81638938–82770
108Azure AI Speech text-to-speechTTSBB75.690–70759030–80860
109Amazon SESEmailBB75.575–85648945–85870
110Amazon TranscribeSTTBB75.575–90808345–60850
111ElevenLabs Scribe Speech to Text APISTTBB75.580–95756560–77710
112OpenAI embeddingsEmbeddingsBB75.565–89909545–6090-2
113Klaviyo API + MCPLifecycle marketingBB75.477–87788438–77800
114ZoomInfo API + MCPLeadsBB75.487–84716560–83760
115SuprSendNotificationsBB75.469–94856660–81730
116Telnyx API + MCPMessagingBB75.480–93904566–83720
117CosmicCMSBB75.490–88824768–86630
118SanityCMSBB75.287–87746760–8987-3
119OpenAI Agents APIHarnessesBB75.260–88808745–87850
120Cohere Embed and RerankEmbeddingsBB7573–92875560–90720
121gooseHarnessesBB7577–81827790–8667-5
122Neon MCP ServerDatabasesBB7572–90796960–80750
123StatsigProduct analyticsBB74.979–82717460–81760
124ElevenLabs Agents API + MCPVoice agentsBB74.975–92778353–8578-3
125GitHub MCP ServerCodeBB74.965–83788460–9384-2
126CodexDataBB74.985–91794976–80590
127Big ArrowHuman approvalBB74.875–78776790–70630
128Grafana MCP ServerObservabilityBB74.891–81746790–9680-7
129AblyWebhooksBB74.787–80876460–8580-3
130Pulumi ESCSecretsBB74.760–82757983–82660
131Deepgram Voice Agent APIVoice agentsBB74.775–90677160–85780
132Scalekit AgentKitAgent authBB74.772–87886660–80660
133WooCommerce API + MCPCommerceBB74.780–77835590–8276-2
134GeocodioMapsBB74.770–92805860–86870
135NocoDBSpreadsheetsBB74.697–85747145–8777-3
136Nylas Notetaker APIMeetingsBB74.678–89777860–8177-3
137SeleniumBrowserBB74.690–89704390–9477-3
138Amazon SNSNotificationsBB74.575–83658953–65860
139Google Cloud Model ArmorGuardrailsBB74.490–78759030–8587-3
140Telnyx Voice API + MCPCallingBB74.475–93904566–83720
141Firebase Cloud MessagingNotificationsBB74.476–86706860–83820
142GraphicleMemoryBB74.463–83856790–8371-2
143PuppeteerBrowserBB74.480–81675290–89710
144Context7CodeBB74.468–86756990–9568-3
145SpiderScrapersBB74.385–87704294–75700
146Upstash QStashWebhooksBB74.383–78836160–77760
147Langfuse API + MCPEvalsBB74.380–93747760–8886-5
148AXLRHarnessesBB74.370–80836290–79550
149Intercom API + MCPSupportBB74.288–94747445–7780-3
150LinqMessagingBB74.283–93905060–8372-2
151Nutrient DWS Processor APIPDFBB74.178–82796860–8376-1
152LangSmith API + MCPEvalsBB74.181–86767960–8578-4
153Massive (formerly Polygon.io)MarketsBB74.178–89783598–82670
154OpenHandsHarnessesBB74.190–90786490–8567-7
155DBHubDatabasesBB74.180–73854390–95610
156Circle Wallets (Agent Wallets, Programmable Wallets)WalletsBB7460–88756588–77730
157AgnoFrameworksBB7483–90766475–8684-5
158Azure Key VaultSecretsBB7487–85758630–8078-2
159RestateHuman approvalBB7496–85816160–9253-3
160OpenAI Moderation APIGuardrailsBB7465–92859545–4790-2
161Yield.xyz MCP serverWalletsBB7465–958256100–60530
162Respan API + MCPEvalsBB73.983–89907445–8559-3
163LaunchDarklyProduct analyticsBB73.874–89696660–82810
164Opik API + MCPEvalsBB73.872–84935753–85750
165LlamaIndex FrameworkFrameworksBB73.870–85737490–6967-2
1661Password service accounts, SDKs and Environments MCPSecretsBB73.768–74699445–82870
167Terraform MCP ServerInfraBB73.783–80746590–9086-6
168Drex 1.5DecisionsBB73.680–88857060–6173-2
169FinchPayrollBB73.684–87716645–84750
170OpenAI Realtime APIVoice agentsBB73.662–84768345–82900
171LangGraphFrameworksBB73.675–89706090–9081-4
172OneSignalNotificationsBB73.680–87647453–81730
specUniversal Commerce Protocol (UCP)CheckoutBB73.679–90826763–70470
173SvixWebhooksBB73.585–87866160–8886-5
174Kraken APIExchangesBB73.567–80798060–81650
175AgentMail API + MCPInboxesBB73.461–88866590–8765-3
176Google Cloud Document AIDocumentsBB73.490–81707030–80900
177Privy Wallets (server wallets, agent wallets, policy engine)WalletsBB73.453–83818564–80720
178ImpeccableDesignBB73.485–78665090–79690
179BrowserbaseBrowserBB73.390–80754398–9272-4
180MailerSendEmailBB73.395–80696345–85680
181Vercel ConnectAgent authBB73.373–84758360–8177-3
182Zoho CRMCRMBB73.372–81816760–77750
183dYdXExchangesBB73.383–58667890–83500
184Google Forms APIFormsBB73.291–78726653–61810
185LokaliseTranslationBB73.284–76697345–84810
186Azure TranslatorTranslationBB73.283–83827530–70780
187E2BSandboxesBB73.277–92706275–9370-3
188NVIDIA NeMo GuardrailsGuardrailsBB73.275–79676290–70720
189AsanaProjectsBB73.177–91716545–80840
190EasyshipShippingBB73.180–87736160–76720
191Merge Agent HandlerTool accessBB73.159–80688383–69720
192RemoteHRBB73.196–78577453–72690
193SteelBrowserBB73.163–86795688–82650
194Bandwidth Messaging API + MCPMessagingBB73.175–86736560–78750
195InngestHuman approvalBB73.172–89767460–8854-1
196NitrosendLifecycle marketingBB73.148–93836283–81750
197Spree CommerceCommerceBB72.984–87906583–9164-8
198ZepMemoryBB72.970–93878445–8061-3
199MCP GatewayTool accessBB72.987–75778045–9060-2
200Netlify API + MCPInfraBB72.967–79695983–87810
201YoutrustE-signaturesBB72.985–90636860–61720
202xAI API (Grok)ModelsBB72.955–88909730–77630
203AirtableSpreadsheetsBB72.894–73687845–49860
204Claude Agent SDKFrameworksBB72.870–89958360–8783-8
205DeciderDecisionsBB72.890–78803890–76520
206Apideck Accounting API + MCPAccountingBB72.878–92906645–8074-3
207Pipedream API + MCPWorkflowsBB72.882–74736660–79750
208BoldSignE-signaturesBB72.884–85626360–78770
209Groq Speech-to-TextSTTBB72.690–58757960–48850
210monday.comProjectsBB72.679–88827045–8579-3
211OpenCodeHarnessesBB72.672–88796090–8171-4
212fal music modelsMusicBB72.677–89837330–82640
213ClickHouse MCP ServerDatabasesBB72.574–79656690–8482-3
214VespaRetrievalBB72.580–58807560–9186-2
215Wikimedia REST APIDataBB72.567–68686690–70920
216GraphifyCodeBB72.558–77776690–8081-1
217Agent Development Kit (ADK)FrameworksBB72.480–87809290–9069-12
218Amazon Bedrock AgentCore Code InterpreterSandboxesBB72.485–81758745–5573-2
219Nebius AI CloudGPU computeBB72.477–78778130–82770
220Qwen CodeHarnessesBB72.469–91855390–8866-4
221LayaDecisionsBB72.465–80805790–73660
222Merge Accounting APIAccountingBB72.475–86726753–83700
223REACodeBB72.377–86735790–8859-3
224WebflowCMSBB72.384–87727445–8081-3
225past.devMemoryBB72.368–86836460–77660
226Cloudflare MCP ServersInfraBB72.367–78797668–7988-3
227Gladia Speech-to-Text API + MCPSTTBB72.370–95756560–70650
228VendureCommerceBB72.389–91686575–8574-6
229Qdrant API + MCPRetrievalBB72.270–87878345–8584-5
230DeelHRBB72.288–85657145–70690
231ZeroClawAssistantsBB72.274–82727690–9068-6
232Zyte APIScrapersBB72.285–85725560–77670
233fal video modelsVideoBB72.277–89817330–81640
234Azure MCP ServerInfraBB72.169–77747390–8980-5
235ValyuSearchBB72.170–90834468–80780
236Murf TTS API + MCPTTSBB7290–85716860–7367-3
237fal image modelsImageBB7277–91807330–79610
238Axiom MCP ServerObservabilityBB7272–62637290–83700
239letmeTool accessBB71.975–89795590–5579-3
240GPT-LiveVoice agentsBB71.962–87768130–82900
241DopplerSecretsBB71.980–81598245–76750
242EntireCodeBB71.961–84766660–83840
243EverOSMemoryBB71.979–89784760–80710
244Twelve DataMarketsBB71.983–83845445–82670
245Avalara AvaTaxTaxBB71.867–85846545–79800
246Fireworks AI InferenceModelsBB71.870–93807760–6569-3
247Kontent.aiCMSBB71.882–74718130–83760
248PubNubWebhooksBB71.881–77716360–83660
249pplx-deciderDecisionsBB71.875–89836838–65740
250Mailgun API + MCPEmailBB71.872–83607460–87700
251Soniox Text-to-SpeechTTSBB71.887–85788730–6373-3
252Celesto (formerly SmolVM)SandboxesBB71.866–85695290–81700
253Perplexity Search APISearchBB71.880–89855638–82660
254OpenAI GuardrailsGuardrailsBB71.773–66735990–8982-2
255OpenfortWalletsBB71.788–89656571–8771-5
256ShopwareCommerceBB71.783–85787375–8778-8
257Bandwidth Voice API + MCPCallingBB71.773–88657053–78750
258Coda (Superhuman Docs)SpreadsheetsBB71.691–76617145–72780
259Cloudflare Sandbox SDKSandboxesBB71.690–77637345–9086-3
260CrowdinTranslationBB71.578–80696845–82800
261LandingAI Agentic Document ExtractionDocumentsBB71.585–88754360–85640
262Coinbase Developer Platform (Agentic Wallet, AgentKit, CDP MCP)WalletsBB71.562–94837890–7777-8
263Google Weather API (Maps Platform)WeatherBB71.585–82797360–20720
264Zernio (formerly Late) API + MCPSocialBB71.580–83886560–8771-5
265withHumanHuman approvalBB71.556–91868345–7875-2
266MoodleLearningBB71.480–66496790–81760
267EmailEngineMailboxesBB71.379–91817160–8175-6
268Duffel Flights and Stays APITravelBB71.380–59806060–92760
269Kilo Code CLIHarnessesBB71.285–90726475–8776-7
270AmplitudeProduct analyticsBB71.175–72727445–83790
271GrowthBookProduct analyticsBB71.186–76767660–8975-6
272OpenSearchRetrievalBB71.177–77807090–8461-6
273CoinMarketCap x402 APIDatabasesBB71.177–72815596–72330
274ElevenLabs Music APIMusicBB71.180–90706530–82780
275Google Cloud TranslationTranslationBB71.190–70748730–38850
276Hermes AgentAssistantsBB7157–80807590–8467-4
277Calendly API + MCPSchedulingBB7185–86617045–60800
278DeepL APITranslationBB7152–95838245–8781-3
279AembitAgent authBB70.965–85758460–8362-3
280BankMCPBank dataBB70.963–75706990–80500
281Circle Gateway NanopaymentsPlatformsBB70.866–80756489–7773-3
282ClairicDocumentsBB70.858–85787560–57810
283StrapiCMSBB70.882–80656675–8772-4
284DaytonaSandboxesBB70.885–87557375–8064-4
285Infobip API + MCPMessagingBB70.872–90667438–70820
286commercetoolsCommerceBB70.778–86837030–8582-3
287OpenAI Image APIImageBB70.765–90728530–7990-2
288Cloudflare Workers AIModelsBB70.668–80758063–7276-3
289PlanetScale MCP ServerDatabasesBB70.678–82747723–80720
290Square Appointments (Bookings API)BookingsBB70.655–89807060–71710
291Google VeoVideoBB70.687–71648030–77770
292Twilio SendGridEmailBB70.676–79608145–67820
293KevDecisionsBB70.573–77785290–63550
294BrowserlessBrowserBB70.482–89785360–8564-3
295Vela 2.0DecisionsBB70.457–78796390–74540
296Speechmatics Speech-to-TextSTTBB70.370–90806560–7076-3
297CLI ManagerHarnessesBB70.268–87655890–62580
298ShotstackAssetsBB70.295–88634868–7373-3
299Harness Feature Management & ExperimentationProduct analyticsBB70.280–71837923–81610
300LiveKit TelephonyCallingBB70.165–84666160–90760
301Front API + MCPSupportBB70.177–81677345–60800
302SerpKiteSearchBB70.175–90973760–8077-3
303Bria APIImageBB7083–84765360–69520
304Soniox Voice CloningCloningBB7087–86826730–7571-3
305VirusTotalSecurity intelBB7055–77836960–88650
306Cerebras InferenceModelsB69.970–93857030–65600
307MetronomePlatformsB69.971–88756859–8266-3
308MiniMax CodeHarnessesB69.951–81795790–79630
309PostHogProduct analyticsB69.984–81788460–8983-10
310Smartsheet API + MCPSpreadsheetsB69.985–88746630–8083-3
311SquareCommerceB69.953–87836760–71710
312BigCommerce API + MCPCommerceB69.977–66627060–82770
313CourierNotificationsB69.982–84685960–8674-3
314Open-MeteoDataB69.865–77853875–90750
315OpenCage Geocoding APIMapsB69.885–77864453–8685-3
316Soniox Speech-to-TextSTTB69.877–85708730–7576-3
317Framelink MCP for FigmaDesignB69.868–87835590–41470
318PersonaVerificationB69.770–85767653–7569-3
319email-agent-mcpMailboxesB69.750–79796790–84420
320HygraphCMSB69.683–78726353–64620
321incident.io APIObservabilityB69.690–88756415–8377-2
322GleanKnowledgeB69.660–9380870–85780
323Cloud Run GPUsGPU computeB69.670–81667830–80840
324amCharts 5 MCP ServerCodeB69.665–84705090–75570
325DigitalOcean SpacesStorageB69.587–63737930–67750
326Plaid Identity VerificationVerificationB69.573–93756123–83770
327Bland AI API + MCPVoice agentsB69.565–68537985–70720
328CoinGecko x402 APIDataB69.560–72616094–82740
329CirclebackMeetingsB69.577–81726145–78700
330PreludeMessagingB69.565–93856345–83420
331Render MCP ServerInfraB69.478–78695860–56840
332Cronofy APISchedulingB69.495–58766045–71720
333SignalWire Voice APICallingB69.453–87687260–81740
334Xero API + MCPAccountingB69.468–79756453–76720
335NewsCatcherSearchB69.467–94844545–76790
336Mailtrap Email API + MCPEmailB69.365–82626060–93770
337RobocorpDesktopB69.387–65594490–81650
338TinyFishBrowserB69.360–85796064–80590
339AssemblyAI Speech-to-Text (Universal)STTB69.385–95805060–7977-6
340Zendesk Support APISupportB69.368–83777515–81870
341RightNow System ZeroDecisionsB69.350–83867060–68700
342Canva REST APIs + MCPAssetsB69.279–95596345–8583-3
343Cartesia InkSTTB69.273–89744753–84680
344Cloudflare AI GatewayModelsB69.266–75738063–7574-3
345Vonage Messages API + MCPMessagingB69.277–90507356–80490
346GitLab MCP ServerCodeB69.262–85666645–85870
347Together AI InferenceModelsB69.175–93857030–6574-3
348Zoho DeskSupportB69.175–72676660–63790
349OpenRouterModelsB69.155–90856760–8272-3
350Airwallex Spend and IssuingSpendB6972–69698438–66790
351PhraseTranslationB6970–87705845–83740
352Vercel AI GatewayModelsB6979–74837930–7971-3
353BoxLang AIFrameworksB6966–71747175–8370-3
354MaritimeSandboxesB68.945–87876871–8166-2
355PennylaneAccountingB68.982–85666830–65760
356PipecatVoice agentsB68.968–92726790–8583-9
357MicrolinkScrapersB68.952–89824471–86750
358AI MemoryMemoryB68.865–90606590–50500
359Exa API + MCPSearchB68.877–86964398–8861-9
360PlaidBank dataB68.873–93826723–8379-3
361AxolotlFine-tuningB68.764–80605290–88580
362MailSlurpInboxesB68.775–72847048–55600
363Retell AI API + MCPVoice agentsB68.760–92607660–8577-3
364Sentry MCPObservabilityB68.736–87787560–80810
365ArcGIS Location PlatformMapsB68.670–61786460–83690
366PresidioGuardrailsB68.678–71695390–7867-3
367Gemini EmbeddingEmbeddingsB68.645–89866060–75750
368OlostepScrapersB68.660–88873379–83600
369airSlate SignNowE-signaturesB68.593–81666253–5775-3
370Sinch Voice API + MCPCallingB68.560–85815553–79730
371OpenMausBotHarnessesB68.560–90775975–7459-2
372Azure MapsMapsB68.481–83736630–57750
373HyperbrowserBrowserB68.465–72674499–79660
374Meta Marketing APIAdvertisingB68.475–76776145–77630
375pplx-decider-flashDecisionsB68.475–89816838–6574-3
376LinkupSearchB68.470–88763283–73610
377BerthLocal AIB68.461–69755190–76680
378ContentstackCMSB68.381–79676945–8274-3
379OneTool accessB68.387–77766953–8964-6
380ClefDecisionsB68.368–78757260–5987-3
381GreyNoiseSecurity intelB68.378–83764053–75760
382SuperpowersCodeB68.368–71715890–8368-3
383Cloudflare CLI (cf)InfraB68.248–81657975–6888-2
384RobotomailInboxesB68.253–90726553–77750
385Xweather Weather APIWeatherB68.284–83785248–8169-3
386Mistral AI APIModelsB68.250–93917360–8080-6
387draw.io + MCPDiagramsB68.245–87595590–86810
388Amazon Nova 2 SonicVoice agentsB68.295–75647715–50790
389Keeper Secrets ManagerSecretsB68.178–71638630–9279-3
390LangWatchEvalsB68.163–87677160–9076-4
391SumsubVerificationB68.170–78608038–74740
392TinyFish SearchSearchB68.165–82785755–80600
393Box API + MCPStorageB68.165–91727338–8477-3
394Cloudflare Web Search APISearchB68.177–81637430–71730
395Mistral Embed and Codestral EmbedEmbeddingsB68.150–89956860–5580-3
396DatalabDocumentsB68.158–89735660–79680
397Google Cloud Text-to-SpeechTTSB68.185–80658030–25870
398Azure AI Content Safety (Prompt Shields)GuardrailsB6875–74788623–35830
399NangoAgent authB6878–85747160–9077-8
400DoclingDocumentsB6885–77775390–8965-8
401SpritesSandboxesB6880–74686160–8066-2
402agents-treeEvalsB6860–64736690–63620
403CheckPayrollB67.980–7883748–72630
404Cohere Chat API (Command models)ModelsB67.967–84725553–72750
405Hugging Face Inference EndpointsGPU computeB67.973–73628330–80700
406StackOneTool accessB67.980–78757157–8470-6
407UiPathDesktopB67.982–79738238–8076-6
408agent-browserBrowserB67.948–84806790–7063-3
409Buffer API + MCPSocialB67.975–86655645–73750
410KernelBrowserB67.953–93848173–9061-8
411Unified.to MCP ServerTool accessB67.873–81715372–8467-3
412WhatsApp Business Platform (Cloud API)MessagingB67.875–82676760–34730
413Google Cloud Speech-to-TextSTTB67.875–80708530–25870
414Infobip Calls API + MCPCallingB67.862–90627245–60820
415AbyssaleAssetsB67.841–98855553–85710
416Fireworks AI Fine-tuningFine-tuningB67.755–77758260–8565-3
417Stadia MapsMapsB67.763–86804060–80760
418ArcmiraSearchB67.740–87916183–8461-3
419LemonadeLocal AIB67.675–70703690–76650
420OmnisendLifecycle marketingB67.681–79795745–39770
421Pydantic LogfireEvalsB67.634–81728060–90750
422Braintrust API + MCPEvalsB67.676–85665860–8566-3
423Commerce Layer API + MCPCommerceB67.688–83696438–8259-3
424MarmotKnowledgeB67.660–82846145–8876-2
425Pirate WeatherWeatherB67.675–85774545–80660
426WorkOS Pipes and AgentsAgent authB67.690–63826930–8665-3
427CodatAccountingB67.590–77814915–80730
428Black Forest Labs FLUX APIImageB67.580–88676030–73650
429Playwright MCPBrowserB67.557–70845790–8474-4
430Smallest.ai Pulse STTSTTB67.563–92665260–73710
431MedplumHealthB67.475–82737438–9373-5
432SambaCloudModelsB67.485–81676160–7773-5
433Extend API + MCPDocumentsB67.477–87684545–82700
434Postmark API + MCPEmailB67.460–81796360–8078-3
435TokenEconomizerCodeB67.447–66767190–53760
436Modal SandboxesSandboxesB67.355–79677930–93780
437Lingo.devTranslationB67.381–64756230–87720
438Terra APIHealthB67.367–89786530–75580
439Apollo API + MCPLeadsB67.375–87656045–58730
440MusicGen on ReplicateMusicB67.375–87754045–75780
441NominatimMapsB67.382–51755090–60590
442Google People APIContactsB67.230–84837853–83750
443PhaseSecretsB67.291–58568338–8379-3
444Qwen-Image 2.1 (Model Studio)ImageB67.272–66738030–77650
445FactorialHRB67.188–77586515–84800
446OffscreenSandboxesB67.155–67746290–66610
447Dropbox API + MCPStorageB67.160–91777853–9065-6
448Plain API + MCPSupportB67.165–92686745–8473-3
449YCloudMessagingB67.179–90833153–74530
450Cloudflare Email ServiceEmailB6753–89727945–8177-3
451Gemini EnterpriseKnowledgeB6785–78487715–80820
452StoryblokCMSB6764–68787353–8081-3
453Replicate DeploymentsGPU computeB6775–85734045–78790
454Sinch Messaging APIs + MCPMessagingB6773–81505560–83770
455SendcloudShippingB66.973–89715856–6772-3
456Vapi API + MCPVoice agentsB66.970–89676153–8875-4
457Smallest.ai Lightning TTSTTSB66.958–92725260–68700
458Lakera Guard (Check Point AI Guardrails)GuardrailsB66.870–90838323–21620
459Reducto API + MCPDocumentsB66.877–80844045–80590
460Agent Platform Memory BankMemoryB66.868–75757230–62810
461Decision 3.0DecisionsB66.761–70725390–74510
462Groundmail API + MCPEmailB66.768–88696353–7771-3
463IronClawAssistantsB66.767–74696690–6476-5
464Jina Embeddings and RerankerEmbeddingsB66.780–86883545–62600
465fastCRWScrapersB66.650–91914053–93600
466Arcade.devAgent authB66.563–87867160–7671-7
467LetsFGTravelB66.555–83836685–8865-7
468Nevermined API + MCPPlatformsB66.550–85777774–8173-6
469Upstash Vector API + MCPRetrievalB66.575–49806560–56770
470WeatherAPI.comWeatherB66.580–89803745–60650
471DIDWW Voice and Number Management APICallingB66.471–87775741–8371-3
472MX Platform APIBank dataB66.477–85746023–71640
473NanoClawAssistantsB66.478–82707490–7869-11
474PiHarnessesB66.475–84766190–8767-10
475Alibaba Wan (Model Studio)VideoB66.470–67638030–77780
476Loops API + MCPEmailB66.490–82704645–8366-3
477CensysSecurity intelB66.455–82795553–84640
478DocusignE-signaturesB66.362–70706945–81730
479PlantUMLDiagramsB66.383–65785090–9174-8
480LanceDBRetrievalB66.375–81744430–93750
481Plivo Voice APICallingB66.380–59705060–78710
482CloudConvert APIPDFB66.280–59617153–58750
483LocalAILocal AIB66.284–81716290–8053-9
484LiteLLMModelsB66.270–91797290–8263-12
485SequenzyLifecycle marketingB66.238–95696760–78670
486UserpilotProduct analyticsB66.282–71606245–79590
487Azure Document IntelligenceDocumentsB66.182–86647130–25790
488D2DiagramsB66.168–65784890–7470-3
489Mercury APISpendB66.157–72688245–74630
490Replicate video modelsVideoB66.175–86724145–68760
491jevcrawlScrapersB66.168–81805945–68510
492Notion MCPWorkB66.184–75537145–8173-3
493Plivo APIMessagingB66.190–59575060–78710
494DocuSealE-signaturesB6650–77565283–90800
495neckbeardInfraB6653–75716890–6852-2
496Gemini CLIHarnessesB6671–93786730–8890-7
497Azure DevOps MCP ServerCodeB65.972–64708160–8080-6
498CollibraKnowledgeB65.971–8375680–76850
499KoboldCppLocal AIB65.963–79653890–82550
500Wix Stores and eCommerce APICommerceB65.961–76686145–83770
501PaseoHarnessesB65.950–65705890–75680
502NovuNotificationsB65.898–92846460–9267-15
503GranolaMeetingsB65.870–82717038–6076-2
504Pleo API + MCPSpendB65.881–82557123–64730
505SpecStreamSearchB65.823–80657396–81660
506Uniswap Trading APIExchangesB65.842–90726360–88550
507DeepEvalEvalsB65.759–78724790–8366-3
508MoEngageLifecycle marketingB65.776–8768690–75760
509Spendesk API + MCPSpendB65.770–8762860–62800
510W&B WeaveEvalsB65.768–74716353–8772-3
511YNABPersonal financeB65.782–77715145–8275-3
512Visual Crossing Weather APIWeatherB65.768–74784083–58570
513ConvertAPIPDFB65.754–92795538–80650
514KudosityMessagingB65.785–86733748–57620
515Dropbox SignE-signaturesB65.665–88726538–8569-3
516Mistral Voxtral TranscribeSTTB65.653–85777060–8883-6
517BelvoBank dataB65.584–80806030–30670
518RipplingHRB65.577–73759015–7456-3
519Cursor Agent SDKFrameworksB65.545–82806845–74710
520OKX APIExchangesB65.565–73776830–81620
521ComplyCubeVerificationB65.484–78605241–74630
522Grep AISearchB65.450–83706460–74610
523GustoPayrollB65.463–86826023–76620
524Microsoft Dynamics 365 SalesCRMB65.465–81867438–7976-6
525Oxylabs Web Scraper APIScrapersB65.475–68795360–8264-3
526OnyxKnowledgeB65.472–88777345–7769-7
527Binance APIExchangesB65.445–86767253–81430
528Shop MCPShoppingB65.333–89797483–6661-3
529SignWellE-signaturesB65.387–75524753–77650
530Teamwork.comProjectsB65.373–72685845–9380-3
531SerpApiSearchB65.365–67884045–82840
532BasetenGPU computeB65.290–84558260–9067-11
533IPinfoSecurity intelB65.238–76846060–85700
534PomeradoBrowserB65.140–86836963–8153-2
535Bunny StorageStorageB65.185–67614560–75610
536ClawbooHarnessesB65.152–83757090–5158-4
537TradingAgentsFrameworksB65.152–61616290–78680
538DeepInfraModelsB6575–69776430–61680
539PandaDocE-signaturesB6573–87686145–7759-3
540QwenPawAssistantsB6570–68676090–8358-5
541Lucid API + MCPDiagramsB6580–73628038–33620
542Stytch Connected AppsAgent authB6573–64657630–76660
543FullstoryProduct analyticsB64.965–62776538–70810
544NVIDIA NeMo Retriever Embedding and Reranking NIMsEmbeddingsB64.953–78735560–72720
545Optimizely ExperimentationProduct analyticsB64.983–7668660–78750
546Nansen x402 APIDatabasesB64.945–84864599–7551-3
547Gemini 4 HydrogenModelsB64.945–85905040–85720
548Grounding with Bing SearchSearchB64.968–68617930–80660
549Magic PatternsDesignB64.975–84675535–70600
550OutcrawlScrapersB64.945–71836660–79560
551TodoistProjectsB64.861–76776145–8877-3
552VeriffVerificationB64.867–82436360–74690
553Lusha API + MCPLeadsB64.885–83515960–8280-6
554ScrapeGraphAIScrapersB64.849–887244100–40610
555Strands Decider 2BDecisionsB64.850–76694990–68620
556FLORAImageB64.846–79686853–83680
557ArgylePayrollB64.781–86584823–75830
558FullEnrich API + MCPLeadsB64.770–76665760–48680
559Gemini Developer APIModelsB64.745–85935060–8372-3
560Lara Translate APITranslationB64.780–74774060–8362-3
561LlamaParse API + MCPDocumentsB64.780–82725360–8067-6
562402cronWebhooksB64.745–81764885–62650
563OpenDocRouterDocumentsB64.768–85844145–75490
564ScrapeAtlasScrapersB64.742–88765570–70590
565VoidCodeB64.655–67766675–8854-3
566Filesystem (MCP reference server)DatabasesB64.675–76704590–7076-6
567Make API + MCPWorkflowsB64.673–66636845–56750
568MapTiler CloudMapsB64.643–79706460–70800
569ZeroBounceLeadsB64.675–54696553–70630
570BrexSpendB64.575–80707238–5667-3
571Etsy Open API v3ShoppingB64.582–75646245–45580
572LeadMagic API + MCPLeadsB64.552–93716330–80650
573SMTP2GO API + MCPEmailB64.582–59665960–7375-3
574MemPalaceMemoryB64.577–85755390–7458-9
575MixpanelProduct analyticsB64.465–79617056–8084-5
576ShippoShippingB64.469–83644260–77580
577Crossmint API + Docs MCPPlatformsB64.453–71777344–8580-3
578Vertex AI Gemini tuningFine-tuningB64.467–82486530–80880
579Clerk MCP serverAgent authB64.465–65636560–77570
580Expo Push NotificationsNotificationsB64.475–60634860–84700
581Inworld Instant Voice CloningCloningB64.475–64645560–68630
582Crusoe CloudGPU computeB64.376–79457445–7677-3
583Foundry LocalLocal AIB64.368–53613990–88730
584Decodo Web Data APIScrapersB64.359–67824560–85620
585BasecampProjectsB64.259–80676745–8280-3
586OpenClawAssistantsB64.275–84777490–8672-15
587Brave Search API + MCPSearchB64.180–72725830–7781-3
588Milvus and Zilliz Cloud API + MCPRetrievalB64.195–69626753–8569-8
589Structurizr + MCPDiagramsB64.170–61554375–88760
590Higgsfield APIVideoB6480–70814830–65640
591PrefactorEvalsB6453–82885660–8249-3
592Figma API + MCPDesignB6464–86607115–84670
593Miro API + MCPDesignB6473–85536745–6777-3
594Mistral OCR APIDocumentsB6453–89837060–5876-6
595Slack MCP Server (official)WorkB6473–64517930–68820
596Tencent Hy4ModelsB6446–73867145–73520
597Socure RiskOSVerificationB63.965–84466053–75690
598Wasabi Hot Cloud StorageStorageB63.965–59688360–4671-2
599ModalGPU computeB63.970–70576830–85710
600Gemini 4 CarbonModelsB63.940–85935530–83720
601WebScraping.AIScrapersB63.943–89953660–82510
602Fetch (MCP reference server)SearchB63.855–70934790–7275-6
603GoVisit ToolsTravelB63.845–87734783–76400
604Mailchimp Transactional (Mandrill)EmailB63.870–79606853–28710
605UnipileMailboxesB63.872–77734260–7972-3
606DigitalOcean MCP ServerInfraB63.767–66676750–7087-3
607PlaneProjectsB63.789–80786845–8076-11
608Cartesia Sonic TTS API + MCPTTSB63.773–86654745–8473-3
609Replicate image modelsImageB63.775–83663545–70750
610TrueLayerBank dataB63.771–89736615–48630
611FlagsmithProduct analyticsB63.765–67556560–8377-2
612Fireflies.aiMeetingsB63.660–80685145–80680
613MatomoProduct analyticsB63.638–73787145–79740
614SandBaseTool accessB63.663–76814560–68480
615SmartRecruitersRecruitingB63.595–7551655–58790
616TokenripMemoryB63.557–93707175–7571-9
617LettaHarnessesB63.558–70673590–85550
618OpenBBDataB63.568–73634690–5268-2
619HarnessHarnessesB63.440–85775460–75640
620OneDrive and SharePoint files (Microsoft Graph)StorageB63.474–89847130–7676-9
621Fish Audio Voice Cloning APICloningB63.475–81733545–72610
622Trigger.devHuman approvalB63.375–91797560–9082-15
623Fish Audio TTS APITTSB63.377–84673545–72610
624Pipedrive API + MCPCRMB63.358–80615345–83760
625OpenPanelProduct analyticsB63.370–75544153–83830
626DustKnowledgeB63.253–75677530–86590
627SeaTableSpreadsheetsB63.259–81745945–7676-3
628ShortcutProjectsB63.269–75575445–73740
629MiniMax Video APIVideoB63.275–96763830–57560
630Runloop DevboxesSandboxesB63.260–85666075–8352-5
631Voyage AI embeddings and rerankersEmbeddingsB63.245–61985560–78500
632Gemini Decision 1DecisionsB63.240–80854560–79700
633LumboxInboxesB63.245–82787160–23720
634AidboxHealthB63.159–77797438–8048-3
635Cisco AI Defense Inspection APIGuardrailsB63.190–5967810–8074-3
636Granite GuardianGuardrailsB63.156–60695890–38710
637Mistral Moderation APIGuardrailsB63.143–88757060–8380-6
638Basedash MCPKnowledgeB63.170–57556848–75740
639Node-REDWorkflowsB6380–41445190–81680
640Coinbase Advanced Trade APIExchangesB6358–80736830–57670
641Pipecat PhoneLLMVoice agentsB62.958–72616960–58580
642nanobotAssistantsB62.970–72746390–8254-9
643Gauge CLIEvalsB62.960–60755560–82530
644New Relic MCP ServerObservabilityB62.960–51597645–80810
645Attio API + MCPCRMB62.877–82625645–6369-3
646AffinityCRMB62.759–91666615–70670
647WorkableRecruitingB62.783–67556645–35690
648mini-swe-agentHarnessesB62.765–58695890–32560
649Diagrams.so API + MCPDiagramsB62.745–85856745–8176-5
650AnrokTaxB62.688–71705923–47550
651BILLSpendB62.681–83605638–32670
652BasiqBank dataB62.675–84625530–48720
653ClickUpProjectsB62.662–79576445–54750
654DataForSEOSEOB62.659–86844560–8276-6
655ShipBobShippingB62.675–85565530–51780
656WordPressCMSB62.678–65746290–8369-11
657Help Scout API + MCPSupportB62.675–59586345–61750
658Memory (MCP reference server)DatabasesB62.675–70673990–7576-6
659Gradium Voice Cloning and Voice DesignCloningB62.663–83674845–70630
660Honeycomb MCPObservabilityB62.670–57536845–81700
661Inworld Realtime APIVoice agentsB62.662–60685860–71630
662Granite Switch 4.1DecisionsB62.546–57665490–68760
663HatchetHuman approvalB62.574–85826260–9072-12
664ScrapewiseScrapersB62.545–79725660–79540
665Assymetrix MCP ServerDataB62.536–81916753–5571-2
666Sourcegraph MCP ServerCodeB62.560–71727623–8378-3
667smolagentsFrameworksB62.564–81686690–3465-6
668AshbyRecruitingB62.481–82596315–6478-2
669BrazeLifecycle marketingB62.473–68507715–69830
670Meeting BaaSMeetingsB62.462–93734660–8569-6
671Mirage (AgentVFS)DatabasesB62.446–68795790–8344-3
672FreestyleSandboxesB62.447–86695368–71440
673Kiro CLIHarnessesB62.357–79676960–7371-5
674Hook0WebhooksB62.265–79656460–7784-7
675NotteBrowserB62.270–74755660–8362-6
676SendblueMessagingB62.261–66764660–72580
677meteoblue Weather APIWeatherB62.275–77665151–37620
678LocationIQMapsB62.283–78745545–3610
679Google Search Console APISEOB62.240–75825660–71590
680Rime PrismVoice agentsB62.255–69735660–7572-2
681Mistral Vibe CLIHarnessesB62.169–85845890–7381-14
682TrelloProjectsB62.183–63526245–38800
683Coresignal API + MCPLeadsB62.160–86735830–8271-3
684JevDecisionsB62.148–90865330–66580
685Adyen MCP serverPlatformsB6262–73566353–42830
686Anchor BrowserBrowserB6277–80735445–5753-3
687Recall.aiMeetingsB6250–80715453–63670
688TeableSpreadsheetsB6272–75656245–7677-5
689HonchoMemoryB6250–84715283–8369-6
690ScrapingBeeScrapersB6275–60683760–75650
691CoreWeaveGPU computeC61.845–79587430–80780
692Kimi Code CLIHarnessesC61.868–85714590–7262-8
693Qoder CLIHarnessesC61.851–81805853–7261-3
694Bannerbear API + MCPAssetsC61.845–83715753–67610
695Blaxel SandboxesSandboxesC61.858–80706460–8565-6
696Eden AIModelsC61.833–81758245–58550
697Mailchimp Marketing APILifecycle marketingC61.872–76734638–6874-2
698PDFMonkeyPDFC61.860–61753660–81780
699DocumensoE-signaturesC61.785–79645245–9375-8
700Ultravox Realtime APIVoice agentsC61.752–84694660–53710
701Tencent Cloud TokenHubModelsC61.746–73837153–7354-3
702Truto Unified Calendar APISchedulingC61.777–75715230–50610
703VouchedVerificationC61.763–81663945–70760
704DirectusCMSC61.682–81736883–8662-15
705HimalayaMailboxesC61.684–70654390–7872-9
706Nemotron 3.5 Super VLModelsC61.658–74655453–71580
707Prisma AIRS AI Runtime Security APIGuardrailsC61.677–7365850–6670-3
708Dropcontact API + MCPLeadsC61.685–54645960–15720
709Cronitor MCP serverObservabilityC61.655–80735860–55370
710DatabentoMarketsC61.650–75806133–84490
711QuadernoTaxC61.591–77563945–28770
712CrewAIFrameworksC61.575–74806560–8879-12
713Mailjet API + MCPEmailC61.562–76535245–79740
714Supermemory API + MCPMemoryC61.555–88766445–8552-5
715Vibe-TradingHarnessesC61.563–72677090–8253-9
716InboundInboxesC61.473–76786523–6357-3
717Zapier MCP (agent actions)Tool accessC61.460–54675845–81780
718RelaticleCRMC61.440–86595745–90730
719WoosmapMapsC61.378–81764345–26540
720KintsugiTaxC61.271–87755220–7860-3
721LagoPlatformsC61.286–87713775–9163-11
722ketteiDecisionsC61.230–82896853–57440
723Enrich Layer API + MCPLeadsC61.277–79684060–26590
724Vonage Voice API + MCPCallingC61.245–88607353–51490
725MagicBellNotificationsC61.280–68635053–39590
726RampSpendC61.241–86667938–7471-3
727BeliefStateMarketsC61.120–86675984–78520
728EODHDMarketsC61.165–70763945–82540
729Financial DatasetsMarketsC61.143–80734969–56630
730PreplyLearningC61.155–60635090–58580
731SpAItial API (Echo)3D worldsC61.145–81826860–4065-3
specAgentic Commerce Protocol (ACP)CheckoutC6154–90775363–16520
732Templated API + MCPAssetsC6165–74704253–8076-3
733VoximplantCallingC6185–71456038–7572-3
734ZooCADC6155–65446753–90690
735Mirror CLIHarnessesC60.940–75585890–47660
736CheramiInboxesC60.940–85804653–74560
737Hunter API + MCPLeadsC60.950–90544360–63790
738JanLocal AIC60.961–56464690–77700
739Microsoft Graph Calendar APISchedulingC60.965–93887253–7676-14
740Bitfinex APIExchangesC60.985–48616260–41490
741Prism InferenceModelsC60.865–82686830–4963-2
742Inworld Realtime TTS-2TTSC60.862–60686160–6863-2
743LearnosityLearningC60.875–67615415–85710
744Hostinger Agentic MailInboxesC60.742–84767015–77620
745Seedance on BytePlus ModelArkVideoC60.740–61857930–65650
746Prime AgentHarnessesC60.658–45735090–7967-3
747Veryfi API + MCPDocumentsC60.670–50754460–68590
748DeribitExchangesC60.645–88746238–65460
749CentrifugoWebhooksC60.586–81783790–8660-13
750LinkedIn Marketing APIsAdvertisingC60.566–66635245–64690
751People Data LabsLeadsC60.573–70774338–7966-3
752screenpipeLocal AIC60.553–83755560–6969-5
753MemMachineMemoryC60.551–79654160–83570
754Chatway MCPSupportC60.440–64767045–78570
755DeltaBox LiteSandboxesC60.447–68674790–63460
756Egoist AI PassportMemoryC60.430–83758045–40640
757Runway APIVideoC60.467–93755130–6253-3
758AdamCADC60.435–81827423–55720
759ConvoyWebhooksC60.382–78695345–8367-8
760ZulipWorkC60.383–82684245–7776-7
761bundle.socialSocialC60.367–86663453–7866-3
762Medusa API + MCPCommerceC60.355–81724975–9372-8
763OvermindEvalsC60.323–81767045–76650
764RedoxHealthC60.375–61548123–45640
765Ecwid by LightspeedCommerceC60.260–66666123–79720
766Microsoft Advertising APIAdvertisingC60.250–79636045–61650
767AgentpackPackagesC60.247–65644890–56600
768pdfRest API ToolkitPDFC60.258–90523760–70620
769Zoho CPaaS (formerly ZeptoMail)EmailC60.141–78557845–44820
770OpenMetadataKnowledgeC60.184–85856530–8954-12
771folk API + MCPCRMC60.155–91704545–6082-3
772DeepTutorAssistantsC60.178–71634990–8452-9
773MailPaceEmailC60.170–54715038–65750
774BambooHRHRC6045–88616453–6275-3
775ElasticsearchRetrievalC6077–88777253–8372-15
776Outlook Mail (Microsoft Graph)MailboxesC6067–93817153–7676-14
777Geoapify Location Platform + MCPMapsC6060–84722945–69670
778SkydiveAssistantsC6065–81605953–5351-2
779Zus HealthHealthC6081–7767498–64600
780dotCMSCMSC6047–65686353–75540
781Weaviate API + MCPRetrievalC59.935–84958060–9073-12
782OrcaHarnessesC59.958–65704360–75540
783TruliooVerificationC59.927–82747930–75570
784GreenhouseRecruitingC59.876–7963850–6579-6
785GumloopWorkflowsC59.872–76665030–7758-2
786BybitExchangesC59.845–73666060–66510
787MeetStreamMeetingsC59.845–83884260–8257-4
788PayPal MCP serverPlatformsC59.847–70664860–64780
789Payhawk API + MCPSpendC59.860–71527423–58790
790ShuftiVerificationC59.865–58476253–72680
791Coinbase MCP server (Coinbase for Agents)ExchangesC59.755–65667041–53620
792Microsoft Graph People and Contacts APIContactsC59.760–93837653–7675-14
793TallyFormsC59.675–82504745–6675-3
794Zoho BooksAccountingC59.656–71516845–51750
795Instructure CanvasLearningC59.563–7763610–72820
796PrintFriendly MCP serverPDFC59.542–74874660–58490
797Bright DataScrapersC59.440–74796960–8761-6
specL402Pay per callC59.455–65615896–4753-3
798Mindee APIDocumentsC59.480–82684038–8567-6
799Tray.ai API + MCPWorkflowsC59.472–80336630–60690
800Together GPU ClustersGPU computeC59.438–92556830–75650
801Activepieces API + MCPWorkflowsC59.365–83706460–8077-11
802tldraw SDK + MCPDiagramsC59.370–79714053–8951-5
803Microsoft Foundry fine-tuning (Azure OpenAI)Fine-tuningC59.277–77478530–5582-7
804OpenWeather One Call APIWeatherC59.245–68734290–40590
805LayerreAssetsC59.242–87694838–78630
806Smallest.ai Instant Voice CloneCloningC59.265–70624830–72720
807MattermostWorkC59.183–76636375–8977-15
808FreeAgent APIAccountingC59.185–50554445–60740
809Health GorillaHealthC59.155–8072668–63610
810BreakcoldCRMC5945–90676023–60670
811KestraWorkflowsC58.989–82734175–9371-15
812SPACESandboxesC58.932–75666490–11700
813Cartesia Voice Cloning API + MCPCloningC58.973–85654215–8473-3
814Laminar API + MCPEvalsC58.952–88804445–8364-5
815SavvyCal API + MCPSchedulingC58.970–74535045–40730
816TaskadeProjectsC58.940–91684530–85690
817TraycerHarnessesC58.965–65654245–75600
818Webhook RelayWebhooksC58.943–71596175–7563-3
819Warp (Warp Agent CLI and the Oz Agent API)HarnessesC58.852–89665230–8469-3
820LM StudioLocal AIC58.836–56695990–7262-2
821EasyPostShippingC58.868–55634733–85710
822LearnHarnessAssistantsC58.849–63605290–7360-3
823TigrisStorageC58.772–54535760–7765-3
824ntfyNotificationsC58.783–53663875–7780-7
825Aura AutopilotDesktopC58.741–74695375–43550
826BxPlaywrightBrowserC58.741–66634390–53700
827GateExchangesC58.742–70666053–77490
828LightpandaBrowserC58.768–82754660–7365-8
829Redis MCPDatabasesC58.752–64395490–47770
830X Ads APIAdvertisingC58.775–64624830–63630
831NitroStackFrameworksC58.645–85805560–4543-2
832Salesforce DX MCP ServerCodeC58.660–70535790–2670-3
833DiffbotScrapersC58.655–83632160–77680
834SkyvernBrowserC58.567–88765924–8576-9
835GitHub Copilot CLIHarnessesC58.555–72726560–7774-8
836PusharyHuman approvalC58.542–83866615–6376-3
837DevinHarnessesC58.535–80627230–63750
838HERE Location ServicesMapsC58.455–8993650–4563-3
839NeverApplyRecruitingC58.445–57666668–50570
840WhoisXML APISecurity intelC58.465–70653938–76600
841PrismicCMSC58.373–60635053–7973-5
842ClayLeadsC58.340–84664845–75610
843MEXC APIExchangesC58.343–64646547–75580
844Zoom Realtime Media Streams (RTMS)MeetingsC58.367–7159630–73730
845PinpointRecruitingC58.264–7773780–4078-3
846Pylon API + MCPSupportC58.285–78476115–6268-3
847Salt Edge Account InformationBank dataC58.257–77675115–62820
848Twenty API + MCPCRMC58.252–88806545–9780-12
849PictifyAssetsC58.245–88744560–6852-3
850PrismaticWorkflowsC58.267–7563540–83650
851ScrapelessScrapersC58.170–66503753–81580
852UXKINDesignC58.145–74676818–76620
853Microsoft Power AutomateWorkflowsC5860–68696138–7676-5
854Close API + MCPCRMC5856–82566645–4968-3
855ETG API (RateHawk)TravelC57.965–85694115–65590
856opencommerceShoppingC57.925–81696860–7563-3
857TinkBank dataC57.872–61696315–26810
858Workato API + MCPWorkflowsC57.865–61556845–6070-3
859AmpHarnessesC57.823–72695753–79770
860Droid CLIHarnessesC57.854–79726315–7965-3
861TaxJarTaxC57.884–48614345–44720
862DreamlitLifecycle marketingC57.751–59677415–80590
863FastioStorageC57.732–65737745–7064-2
864Vast.aiGPU computeC57.650–78577030–8164-3
865AWS MCP ServersInfraC57.648–69748690–7975-15
866AbuseIPDBSecurity intelC57.675–59743745–38630
867Docker Model RunnerLocal AIC57.675–49584090–8574-7
868EmdashHarnessesC57.655–55553890–70540
869Hy MemoryMemoryC57.565–55723790–7663-6
870Bitwarden Secrets ManagerSecretsC57.566–53527638–36690
871MiddeskVerificationC57.562–82565615–64620
872Muse CodeHarnessesC57.445–84796730–7556-5
873BeamGPU computeC57.455–58586530–8580-2
874Permit MCP GatewayHuman approvalC57.355–53838030–28460
875Rime TTS API + MCPTTSC57.347–59605960–7076-2
876Kagi Search APISearchC57.343–71756430–41740
877ShieldGemmaGuardrailsC57.346–56725378–40540
878FathomMeetingsC57.258–80644245–37680
879MonidTool accessC57.247–57665090–6646-2
880PlaceCallVoice agentsC57.245–57816075–7451-5
881Atlassian Rovo MCP ServerWorkC57.250–48738515–8078-3
882UnslothFine-tuningC57.248–71534490–8238-2
883Inworld STTSTTC57.260–54566360–41600
884TickerLayerMarketsC57.270–88794998–7938-15
885guilds.runFrameworksC57.243–58676890–2160-2
886Context Mode EngineMemoryC57.160–67803560–5554-2
887Odoo External APIAccountingC57.144–51676133–73910
888PaidPlatformsC57.135–64706545–77520
889Gorgias API + MCPSupportC5775–53475653–27800
890SwellCommerceC5767–71654238–8260-3
891AntseedModelsC5752–55564593–7959-3
892IBM Watson Speech to TextSTTC5740–55666860–33800
893OnshapeCADC5760–63466333–79600
894RunpodGPU computeC56.957–81626730–8264-6
895PayFitPayrollC56.985–78515115–18750
896PrivateGPTLocal AIC56.970–83643290–4164-7
897Capsule CRMCRMC56.883–48594645–28760
898Fastmail API (JMAP)MailboxesC56.859–52736445–16740
899SculptorHarnessesC56.843–45505790–75570
900TableCheck APIBookingsC56.883–7772460–25660
901DeepSeek Search and Fetch APISearchC56.772–55814330–42600
902Rev AI Speech-to-Text APISTTC56.775–80524030–40690
903Better StackObservabilityC56.750–46556060–71690
904KrokiDiagramsC56.764–48705690–8670-10
905ROOKHealthC56.775–74493238–73570
906AlationKnowledgeC56.673–6057560–75760
907Sandbox0SandboxesC56.637–77745830–76470
908WrikeProjectsC56.655–78636045–3376-3
909Rutter Accounting APIAccountingC56.670–79673130–62490
910Antigravity CLIHarnessesC56.643–78785345–6855-3
911CowAgentAssistantsC56.674–67594490–8347-9
912Gemini Live APIVoice agentsC56.641–66694860–8372-3
913LocalazyTranslationC56.668–51623753–75570
914NextBillion.aiMapsC56.645–73774630–73580
915RudderStack MCPProduct analyticsC56.645–66526245–85510
916graftCodeC56.664–32643790–63570
917MiniMax APIModelsC56.575–82593030–45650
918Hume EVI (Empathic Voice Interface)Voice agentsC56.555–90662756–35610
919Hume Octave Voice Design and Cloning + MCPCloningC56.558–87752345–45570
920NorthflankGPU computeC56.585–81528260–7056-15
921Kane CLIBrowserC56.550–77655530–77380
922Muse Spark OpenModelsC56.542–61694475–60530
923AssemblyAI Voice Agent APIVoice agentsC56.465–8372550–6468-3
924MailowsInboxesC56.438–70757553–2857-2
925BitgetExchangesC56.440–74836260–7865-8
926MiniMax Voice Clone and Voice DesignCloningC56.475–78563930–38650
927PaperclipHarnessesC56.459–81706490–7866-15
928Vocily AIVoice agentsC56.440–90844660–6148-5
929IterableLifecycle marketingC56.371–6557550–75700
930MiniMax Speech (T2A)TTSC56.375–81563630–38650
931Adobe PDF Services / PDF Extract APIDocumentsC56.250–80575530–43780
932AtlanKnowledgeC56.267–6574750–8070-6
933Agent TreeEvalsC56.233–50674990–60630
934AirtopBrowserC56.269–76654638–14620
935AutoHotkeyDesktopC56.261–54462090–72760
936Kahoot!LearningC56.245–62666923–58730
937KuCoinExchangesC56.245–60625553–73540
938Router.com (Ramp Router)ModelsC56.222–58776660–59640
939Ayrshare API + MCPSocialC56.177–69692630–7771-3
940Novita AIModelsC56.128–62726553–66570
941Elastic Path API + MCPCommerceC5675–8341448–77620
942DatabarLeadsC5628–84664753–73550
943FlinksBank dataC5661–83545038–31610
944Photoroom APIAssetsC5642–86644960–13690
945Sprout Social APISocialC5671–51596130–38710
946Together Code SandboxSandboxesC5632–86704630–83590
947Jais 3ModelsC55.948–54653190–62550
948ClineHarnessesC55.970–77675775–8565-14
949Companies House APIDataC55.940–65675560–35720
950Murf Voice Cloning APICloningC55.983–8463350–40660
951claude-memMemoryC55.975–84745030–7646-8
952The Weather Company Data APIsWeatherC55.875–77544730–36500
953Agent 37 CloudSandboxesC55.757–57604845–65640
954YapilyBank dataC55.755–88773715–6776-3
955ManifoldSEOC55.745–66685460–5460-2
956Railway MCP ServerInfraC55.757–56596560–8374-7
957T3 CodeHarnessesC55.743–55603590–75510
958Upbit Open APIExchangesC55.742–66705923–83540
959KognitosWorkflowsC55.648–71537130–45660
960DocRaptorPDFC55.668–62634153–21690
961Orkes Conductor Human tasksHuman approvalC55.532–65697730–8146-1
962GLiClassDecisionsC55.543–49603890–64650
963Amazon Ads APIAdvertisingC55.445–76624660–6566-3
964IntrinioMarketsC55.483–43703530–70510
965TypeformFormsC55.450–40635645–81700
966Hugging Face Inference ProvidersModelsC55.325–62587930–80710
967Missive API + MCPSupportC55.360–52466145–50770
968PingramNotificationsC55.350–69675160–8154-5
969MLflow TracingEvalsC55.271–78724090–8865-15
970framefieldsVideoC55.261–77744530–8946-5
specAgent Payments Protocol (AP2)CheckoutC55.231–74518465–9580
971Mixpost API + MCPSocialC55.270–76435715–63510
972AgentcardWalletsC55.226–80637163–7468-6
973Rootly MCP ServerObservabilityC55.148–78776530–8181-9
974Zoho RecruitRecruitingC55.156–46646530–48770
975Gemini Omni 1.2 FlashVideoC55.145–69545030–83720
976DataHubKnowledgeC5558–81786515–7764-8
977HashiCorp Vault + Vault MCP ServerSecretsC5566–74648645–7780-15
978PushoverNotificationsC5562–50645245–20870
979Together AI Fine-tuningFine-tuningC5555–83425730–8068-3
980Iris-3BImageC5545–47544490–70520
981Liquid d1DecisionsC54.930–83824153–6057-2
982Snowflake-managed MCP serverDatabasesC54.950–61467753–3584-3
983Claude CodeHarnessesC54.850–80918030–8281-15
984ExcalidrawDiagramsC54.830–65587353–44630
985BaserowSpreadsheetsC54.765–74725460–7265-11
986Linear MCPWorkC54.660–51387545–6576-3
987HyperQwenLocal AIC54.646–56654190–28550
988TencentDB Agent MemoryMemoryC54.651–61683590–7068-6
989PhotonMessagingC54.567–75632930–7857-2
990SocialCannonSocialC54.545–65856245–0530
991VerdaGPU computeC54.560–72636830–8376-9
992MeliusDesignC54.569–70514330–56560
993Apple Ads Platform APIAdvertisingC54.433–69696830–7867-3
994CerebriumGPU computeC54.448–70496045–6564-2
995MLX LMLocal AIC54.456–37542990–8169-1
996Seedream on BytePlus ModelArkImageC54.440–62726930–7362-3
997GhostCMSC54.383–51695675–8575-15
998Semrush API + MCPSEOC54.334–70775426–57680
999Ship24ShippingC54.342–81683445–60550
1000X MCPSocialC54.372–65454030–70590
1001AstrBotAssistantsC54.273–62544490–8142-9
1002PaperformFormsC54.270–69604038–31550
1003Beam 1.0ModelsC54.152–84674515–50570
1004Framer Server APIDesignC54.143–69474938–77750
1005Git (MCP reference server)CodeC54.155–61714390–7275-10
1006Jina ReaderScrapersC54.170–77803253–4562-7
1007Lambda CloudGPU computeC54.155–69637030–5600
1008GladlySupportC54.175–8442430–68590
1009Microsoft MCP Server for EnterpriseKnowledgeC54.153–29558550–43570
1010Jam MCPCodeC5445–55506360–60480
1011MeetGeekMeetingsC5470–70593845–10670
1012PostNitroAssetsD53.945–77685630–6558-3
1013GraphitiMemoryD53.950–71512790–6573-4
1014AgentzonCommerceD53.940–83784560–5450
1015Microsoft Excel (Microsoft Graph workbook API)SpreadsheetsD53.969–85736323–3376-9
1016RadarMapsD53.970–50655615–7755-2
1017Pusher ChannelsWebhooksD53.863–35455245–76780
1018TinkerFine-tuningD53.735–70535730–87620
1019FormbricksFormsD53.735–74546945–8379-6
1020Prospeo API + MCPLeadsD53.635–61795460–19600
1021Meta Model APIModelsD53.640–60756430–46530
1022Zoom Scheduler APISchedulingD53.665–61435438–38700
1023Ideogram APIImageD53.555–79664330–38490
1024Microsoft Teams (Microsoft Graph)WorkD53.567–92597130–7677-14
1025HyperliquidExchangesD53.540–42706483–39260
1026ShipStation APIShippingD53.570–83643260–3364-6
1027registercheck.de APIDataD53.440–83844045–5841-3
1028InstaVMSandboxesD53.477–66514130–35570
1029sync. labs MCP serverVideoD53.443–81553845–75440
1030CrushHarnessesD53.352–65524183–8259-6
1031Adobe Acrobat SignE-signaturesD53.361–54556638–10710
1032ONLYOFFICE Apps MCP ServerStorageD53.325–73804645–45660
1033Cohere NorthKnowledgeD53.232–8571770–6869-4
1034LlamaFirewallGuardrailsD53.248–49605675–15620
1035PrusaLinkCADD53.255–60483690–17640
1036Refersion MCPAdvertisingD53.272–46505818–58650
1037SmartcatTranslationD53.262–45574038–69740
1038Muse EmberModelsD53.140–60726430–46530
1039Cal.com API v2 + MCPSchedulingD5365–76505645–3580-6
1040HoneyHiveEvalsD5339–86636130–7767-6
1041CurlShotBrowserD5342–63853545–45560
1042PayloadCMSD5376–70645768–7865-15
1043AiderHarnessesD52.944–56655990–1365-4
1044Recraft APIImageD52.965–64623045–38600
1045Kite Agent PassportWalletsD52.945–64687534–7155-6
1046ScrapeOps Proxy API AggregatorScrapersD52.937–67792860–46610
1047SketchDesignD52.953–64703445–35670
1048Weatherbit APIWeatherD52.966–67613945–11630
1049AI Job SearchAssistantsD52.850–55504090–50350
1050Open WebUILocal AID52.860–65576590–8667-15
1051Meteosource Weather MCPWeatherD52.835–72754245–70320
1052Morph CloudSandboxesD52.863–72613930–48440
1053RouterPlusModelsD52.842–59715960–4441-2
1054Tomorrow.io Weather APIWeatherD52.860–75733430–27530
1055Microsoft Copilot Studio MCPWorkflowsD52.746–58627330–4365-2
1056360dialog WhatsApp API + MCPMessagingD52.775–70533553–8490
1057ButterCMSCMSD52.782–78643130–1752-2
1058OctopAssistantsD52.660–60506590–5042-8
1059Chatwoot APISupportD52.668–81475260–7778-12
1060Resemble AI Voice Cloning APICloningD52.565–8166540–3075-3
1061Upload-Post API + MCPSocialD52.557–81723645–7973-9
1062Thunder ComputeGPU computeD52.565–70485130–7966-5
1063AgentWebTool accessD52.430–77774460–20510
1064Crustdata API + MCPLeadsD52.442–8973568–5756-3
1065TaxCloudTaxD52.426–75644551–65530
1066WatiMessagingD52.455–79484030–62520
1067ZampTaxD52.472–44644030–60510
1068DeepSeek Harness (dsh)HarnessesD52.338–77684090–7153-8
1069OutlineWorkD52.367–76627145–8370-15
1070VoiceStudioCloningD52.345–60704545–50490
1071Llama Guard 4GuardrailsD52.338–52675368–28590
1072Breezy HRRecruitingD52.240–79555023–60630
1073ScraperAPIScrapersD52.242–67742960–41540
1074didlogic Voice APICallingD52.226–87574657–38580
1075tl;dvMeetingsD52.250–66734923–28650
1076vLLMLocal AID52.259–68645090–8871-15
1077Datadog MCP ServerObservabilityD52.133–60787630–6869-6
1078NOAA National Weather Service API (api.weather.gov)WeatherD52.125–67664590–8650
1079Whimsical MCPDiagramsD52.160–54465838–31710
1080mails.ai Agent EmailInboxesD52.180–84837845–8065-23
1081ComplyAdvantageVerificationD52.142–82677115–8510
1082SearchAPI.ioSearchD5290–59522845–0620
1083CleverTapLifecycle marketingD5237–57535638–73660
1084HumaansHRD51.965–4460700–36750
1085RecruiteeRecruitingD51.975–44435330–30810
1086ReqRes MCP serverDatabasesD51.930–61763375–50470
1087CreatomateAssetsD51.840–53674853–54520
1088iLoveAPIPDFD51.828–72614145–72630
1089TwitterAPI.io MCPSocialD51.840–62783760–40410
1090RefCodeD51.715–68676353–35690
1091Luma AI APIVideoD51.757–53665330–4767-2
1092Bitstamp APIExchangesD51.763–51586030–28530
1093EvereePayrollD51.787–67534215–6580
1094Magic Hour MCPVideoD51.735–82554523–75580
1095Snapchat Ads APIAdvertisingD51.734–56576530–61700
1096Agent Vault by SelfWalletsD51.640–76667845–1347-4
1097Junie CLIHarnessesD51.633–62555663–6477-4
1098Guardrails AIGuardrailsD51.658–59634490–4461-8
1099Bithumb Open APIExchangesD51.634–64676023–69470
1100CueLightVoice agentsD51.624–50615090–41590
1101HiBobHRD51.640–7847680–59710
1102NutshellCRMD51.645–57445853–36720
1103LangMemMemoryD51.532–55664290–15630
1104Automation AnywhereDesktopD51.535–55606423–60740
1105spoonacularCookingD51.558–65682060–22610
1106Invoice Ninja APIAccountingD51.441–66605353–8363-6
1107OpenDuetAssistantsD51.442–58525945–37680
1108Zoho Mail APIMailboxesD51.442–45566745–33730
1109AgentGateGuardrailsD51.339–47535590–13600
1110KeycardAgent authD51.215–61608645–7945-3
1111AhrefsSEOD51.235–85634535–5867-3
1112PulsediveSecurity intelD51.231–77642375–40590
1113Adobe Firefly APIImageD51.155–7857630–13690
1114Vertex O SeriesTaxD51.160–7451570–32690
1115Celeris-1 DecisionDecisionsD5150–60804530–4455-2
1116LiveAgentSupportD5175–48465245–6968-6
1117Himalayas Remote Jobs MCPRecruitingD50.925–70584275–50460
1118MedblocksHealthD50.928–62726738–29520
1119SPEVDecisionsD50.942–81763330–4959-2
1120AccuWeather Core Weather API + MCPWeatherD50.875–60603938–3510
1121Deno SandboxSandboxesD50.843–66605830–4567-2
1122World Labs World API (Marble)3D worldsD50.742–79704830–10580
1123Financial Modeling PrepMarketsD50.650–61593045–60570
1124RobollyAssetsD50.648–57754753–5490
1125Adobe Commerce (Magento)CommerceD50.555–82676045–7587-15
1126Auggie CLIHarnessesD50.543–67654730–38590
1127Validic InformHealthD50.425–75503060–74670
1128Nebius Token Factory fine-tuningFine-tuningD50.362–6851520–5180-2
1129Paragon ActionKit + MCPWorkflowsD50.370–8041650–3868-4
1130KustomerSupportD50.338–6356648–61660
1131LogMealCookingD50.342–71545830–31570
1132WufooFormsD50.375–33584845–5690
1133CryptoQuant MCP serverDataD50.230–64654560–6559-3
1134LiteAPI (Nuitee Connect)TravelD50.262–86863860–3850-12
1135Zoho PeopleHRD50.257–43486023–48730
1136Chroma API + MCPRetrievalD50.160–66734245–3574-7
1137OllamaLocal AID50.153–79753290–8163-15
1138JumioVerificationD50.140–7345630–75610
1139MusicGPT APIMusicD5040–78543045–59510
1140GoCardless Bank Account DataBank dataD5069–58605715–3560
1141ManatalRecruitingD5075–53434545–6610
1142CogneeMemoryD49.943–85593953–8272-9
1143RomaProjectsD49.940–83604430–5760-3
specACK-Pay (Agent Commerce Kit)Pay per callD49.838–66475884–4362-6
1144HeapProduct analyticsD49.847–57494138–63660
1145HindsightMemoryD49.727–78576545–8050-6
1146Agent ZeroHarnessesD49.757–64625490–7564-15
1147MiniMax Speech-to-TextSTTD49.747–82703330–6630
1148ZeplinDesignD49.536–70554745–33600
1149InboxAPIInboxesD49.533–28615780–43540
1150RocketReach APILeadsD49.564–72723015–11590
1151SurveyMonkeyFormsD49.537–47646245–16710
1152Jerry MCP serverPersonal financeD49.415–65815945–50270
1153NumeralTaxD49.425–87714715–50480
1154HarnessRouterHarnessesD49.335–83736038–7273-11
1155rotki MCP serverAccountingD49.365–76692015–5552-2
1156n8n API + MCPWorkflowsD49.320–92797045–8087-15
1157CookieYes MCP ServerCMSD49.330–46666145–53480
1158Crypto.com Exchange APIExchangesD49.350–42505460–29550
1159OpenProjectProjectsD49.352–76705945–7587-15
1160SpyFu APISEOD49.345–75655230–16380
1161Enable BankingBank dataD49.233–46675230–67630
1162BenzingaMarketsD49.260–7165360–56440
1163TownAssistantsD49.135–33497845–46620
11641upHealthHealthD4960–50326330–27750
1165AutoItDesktopD4959–52471890–17600
1166BreaAssistantsD48.957–37465845–5859-2
1167Copper APICRMD48.979–48402845–16790
1168llama.cppLocal AID48.957–47735290–8164-15
1169GojiberryLeadsD48.970–71553115–20610
1170HelpDeskSupportD48.960–63474453–967-2
1171KarrioShippingD48.976–43574760–2959-6
1172KiteDesignD48.933–62445045–78440
1173SendLayerEmailD48.948–68623830–40490
1174Freshdesk API + MCPSupportD48.840–50484953–26850
1175GuruKnowledgeD48.853–5848570–65610
1176ONESProjectsD48.825–55406545–68620
1177Booking.com Demand APITravelD48.733–9268368–48530
1178Eulerpool Financial Data APIMarketsD48.740–53483560–70520
1179CompanyLensDataD48.639–72704745–3410
1180TeamtailorRecruitingD48.675–2552670–5883-3
1181AtticusAssistantsD48.568–1594990–71570
1182Resemble AI Text-to-Speech APITTSD48.575–7546480–3165-3
1183PocketSmith MCP serverPersonal financeD48.540–46555345–53520
1184FluxA AgentMarketPlatformsD48.412–53665175–64320
1185NexscopeScrapersD48.439–64444153–65410
1186Crisp API + MCPSupportD48.138–46513145–80770
1187Google LyriaMusicD48.142–56406030–7473-3
1188QoviaScrapersD48.140–86776070–7730-15
1189PicoClawAssistantsD4854–51665190–2258-9
1190Microsoft Learn MCP ServerCodeD4820–42586090–13560
1191Windmill API + MCPWorkflowsD4815–79736053–8775-11
1192AutoCuaDesktopD4825–42553890–65420
1193TiingoMarketsD4860–39563245–50570
specVisa Intelligent CommerceCheckoutD47.945–54407034–22600
1194DeepSeek APIModelsD47.960–49703530–4665-3
1195jambonzCallingD47.938–53544060–7852-3
1196KoyebGPU computeD47.850–31515030–73640
1197SugarAI (SugarCRM)CRMD47.861–40695115–12690
1198AttendeeMeetingsD47.726–48613360–70610
1199QuickBooks Online API + MCPAccountingD47.640–38683845–65500
1200TomTom Maps APIs + MCPMapsD47.515–71785138–8658-6
1201AlloyVerificationD47.534–7352490–72630
1202DixaSupportD47.565–73474223–1277-3
1203SimilarwebSEOD47.435–6367528–43620
1204ClickSend SMS API + MCPMessagingD47.465–37703030–6264-3
1205AkoyaBank dataD47.428–70675915–23610
1206Synthflow API + MCPVoice agentsD47.355–6245550–6575-3
1207FreshBooks APIAccountingD47.235–48615745–14650
1208Snipcart API + MCPCommerceD47.268–51402053–38610
1209Puzzle API and MCPAccountingD47.215–65567323–53490
1210Telegram Bot APIMessagingD47.140–54483560–38640
1211Payman Genie MCPPlatformsD47.125–56688034–3046-3
1212Postiz API + MCPSocialD47.143–88654945–8382-15
1213Hexis CloudKnowledgeD47.115–83636630–7551-6
1214Ride with GPS APIHobbiesD47.124–72573045–60590
1215InvofoxDocumentsD4725–70644545–20570
1216JulesHarnessesD4713–76645745–18510
1217FinnhubMarketsD4748–58563245–60270
1218MemoriMemoryD4728–68654568–3352-4
1219Momentic MCP serverCodeD46.945–50335645–53500
1220Crawl4AIScrapersD46.835–55674668–8370-10
1221Moz APISEOD46.742–50743830–28600
1222Functionize MCP serverBrowserD46.655–29435645–39580
1223GAIAAssistantsD46.667–51545715–7862-8
1224Bolna API + MCPVoice agentsD46.540–63524445–7667-6
1225ExcaliClawDiagramsD46.565–37554045–29390
1226BullhornRecruitingD46.450–4764288–76670
1227Travelport TripServices APIsTravelD46.419–84515130–36550
1228TypefullySocialD46.430–58623845–6766-3
1229WoopraProduct analyticsD46.380–45452245–9620
1230ModusToolbox AI AssistantCodeD46.248–20274590–64480
1231LibreTranslateTranslationD46.235–45654560–4062-3
1232gotoHumanHuman approvalD46.220–49624145–77530
1233Fuse MCPLeadsD46.240–50574345–35520
1234Entrust Identity Verification (formerly Onfido)VerificationD46.150–6153380–75500
1235Galileo API + MCPEvalsD46.118–79733330–7455-3
1236BrowserActScrapersD4623–76494138–61440
1237Metriport Medical APIHealthD4635–6570418–44580
1238AnythingLLMLocal AIE45.970–57463890–7862-15
1239Exotel Voice API + MCPCallingE45.960–58404930–5600
1240Vogent APIVoice agentsE45.957–73572623–11570
1241Vaani Developer PlatformVoice agentsE45.928–57644433–59420
1242FluxA AI WalletWalletsE45.820–61605236–66320
1243Instacart Developer Platform API + MCPShoppingE45.822–60616630–20530
1244Strava MCP Connector + APIHobbiesE45.863–5049480–40590
1245pen.devDesignE45.733–36663157–59550
1246Streak API + MCPCRME45.768–55214930–20650
1247Eat This Much Partner APICookingE45.723–7562480–58610
1248Meteomatics Weather APIWeatherE45.734–4180540–52600
1249StepFun APIModelsE45.628–53684130–50570
125046elks Voice APICallingE45.647–66514030–12620
1251TalorysAssistantsE45.538–41326090–6652-6
1252GristSpreadsheetsE45.540–63576345–7863-11
1253LeverRecruitingE45.585–2248538–7710
1254SoldFetchScrapersE45.555–68756260–4941-15
1255atoshipShippingE45.530–55535560–10450
1256Lordevo ShopSEOE45.449–82704597–7660-21
1257SiliconFlowModelsE45.233–65603953–3752-3
1258LeonAssistantsE45.132–33573490–15630
1259AI Diagram Maker MCPDiagramsE45.130–75664530–17350
1260ActiveCampaignLifecycle marketingE45.152–68523330–5600
1261DynaPicturesAssetsE45.142–82452645–4630
1262Harness²HarnessesE4525–54595060–13490
1263Adobe Photoshop APIAssetsE44.955–7536620–2764-3
1264Placid API + MCPAssetsE44.975–37402953–7580
1265Unstructured API + MCPDocumentsE44.927–51584760–8058-6
1266SERPHouseSearchE44.940–66641160–23500
1267Microsoft-Decision-1.1DecisionsE44.850–43505230–44340
1268BlotatoSocialE44.742–82572915–30470
1269FullContactLeadsE44.742–50754315–10650
1270Liner Search MCPSearchE44.750–45653530–45340
1271Ticketmaster Discovery APIHobbiesE44.745–42631845–48650
1272Zo ComputerAssistantsE44.615–73343660–58610
1273Helicone AI Gateway + MCPEvalsE44.650–69705045–5170-13
1274PagerDuty MCP ServerObservabilityE44.635–40596615–17750
1275dsh-superagent (DeepSeek)HarnessesE44.615–61673360–34520
1276Scrape.doScrapersE44.540–57622353–25490
1277JotformFormsE44.538–30605645–10680
1278Nomic EmbedEmbeddingsE44.550–4069620–18500
1279FRED APIDataE44.438–43683560–5540
1280Brevo API + MCPEmailE44.455–58655045–8377-15
1281Skyfire API + MCPPlatformsE44.319–69614945–4550
1282Autodesk Platform ServicesCADE44.352–5952470–53340
1283HTX trading APIExchangesE44.338–40695123–57250
1284PendoProduct analyticsE44.350–2841588–60770
1285PostfleetInboxesE44.335–83726845–7534-15
1286Read AIMeetingsE44.327–46635445–6640
1287Employment Hero PayrollPayrollE44.250–26534715–68610
1288Post Bridge API + MCPSocialE44.142–72604423–3145-3
1289FlightClawTravelE4430–35475075–43320
1290Explorium AgentSourceLeadsE4442–40633260–2861-2
1291GitMCPCodeE4415–56584290–8360
1292Chess.com Published-Data APIHobbiesE43.930–48583575–5540
1293EdamamCookingE43.930–66713038–33310
1294ATXPPlatformsE43.813–54514181–33450
1295fatsecret Platform APICookingE43.635–52645138–13340
1296Mostly Good MetricsProduct analyticsE43.529–49493345–74450
1297Ultravox Voice CloningCloningE43.430–68353560–15660
1298Kiwi.com MCP ServerTravelE43.445–12355483–13660
1299Massed ComputeGPU computeE43.327–71566130–2664-5
1300Rovo Dev CLIHarnessesE43.335–45534823–41630
1301TrackingMoreShippingE43.365–31543135–19570
1302SE Ranking API + MCPSEOE43.151–44582160–4472-5
1303SEC EDGAR APIsDataE42.938–23365290–4570
1304TheFork B2B APIBookingsE42.935–8756428–3548-3
1305RedmineProjectsE42.866–32515190–7352-15
1306SYSTRAN Translate APITranslationE42.725–42545560–4530
1307Mem0 Platform + MCPMemoryE42.625–80634560–7562-13
1308ScrapingAntScrapersE42.535–46722060–5560
1309FilloutFormsE42.530–63413745–16700
1310ContactsPlusContactsE42.428–38427045–5630
1311TextGenLocal AIE42.451–61504090–2455-11
1312Soundverse APIMusicE42.427–75722530–11460
1313FabricCommerceE42.323–7957590–5550
1314PixVerse APIVideoE42.335–58702815–42460
1315Publer API + MCPSocialE42.340–46553923–26680
1316Financial Cents MCPProjectsE42.323–40486138–40510
1317Yodlee Core APIBank dataE42.215–71584923–23550
1318Keywords Everywhere API + MCPSEOE42.142–43544723–10680
1319Microsoft-Decision-1DecisionsE42.140–43505230–34360
1320BitRunLocal AIE4238–30402890–5954-3
1321ShodanSecurity intelE4240–46641960–16440
1322Apiroc Unified Calendar APISchedulingE41.935–44522160–42510
1323Harmony.aiVoice agentsE41.818–77654823–1255-2
specMastercard Agent PayCheckoutE41.840–47387125–13380
1324DragSupportE41.828–71432345–43490
1325FonoaTaxE41.827–6864475–21490
1326Pika APIVideoE41.735–45803330–10460
1327POEditor MCP serverTranslationE41.730–62422745–55410
1328NavigeEvalsE41.628–42615853–3155-5
1329Expedia Group Rapid APITravelE41.625–78693315–15430
1330SimScaleCADE41.615–71554415–33650
1331GPUsMarketGPU computeE41.533–52503045–6950-3
1332TranslatePlusTranslationE41.545–52613545–1052-3
1333Xpoz MCPSocialE41.520–64633745–25290
1334Giggy Speech to TextSTTE41.450–68685745–6930-15
1335Particle HealthHealthE41.428–6953578–7500
1336BotGaugeEvalsE41.318–53555615–6764-3
1337ConductorHarnessesE41.325–6055450–58530
1338Eat App Partner APIBookingsE41.275–28174323–40570
1339PambaSocialE41.245–84753553–7028-15
1340ZammadSupportE40.935–41536660–9571-15
1341PDF.co API + MCPPDFE40.920–74532330–6467-3
1342VerdentHarnessesE40.943–55353330–6155-2
1343HyperFramesVideoE40.845–65602030–5551-5
1344Stormglass APIWeatherE40.834–40654438–5460
1345Alpha VantageMarketsE40.730–46543245–33500
1346Clear Thought 1.5ReasoningE40.728–39533990–443-2
1347ClawpostInboxesE40.68–54534290–2320
1348Vidu APIVideoE40.535–48552830–47460
1349Best Buy Developer APIShoppingE40.431–45723245–5400
1350PowensBank dataE40.425–48575515–13650
1351Epsilla Vector DatabaseRetrievalE40.460–36552745–864-3
1352SecodaKnowledgeE40.340–5944540–10600
1353ScrapingdogScrapersE40.349–5352560–2556-2
1354SubframeDesignE40.219–50474750–6563-5
1355Cursor CLIHarnessesE40.227–39424638–5260-1
1356nTopCADE40.238–6160260–45510
1357DotsAssistantsE40.143–41226115–58420
1358CountlyProduct analyticsE39.923–75666226–8072-15
1359Landbot MCP serverSupportE39.830–50394638–43320
1360HyperbolicGPU computeE39.725–52554823–3964-3
1361MealMe APIShoppingE39.719–72692630–6470
1362Presentations.AI MCP serverDesignE39.715–52604130–55320
1363SalesflareCRME39.718–57494445–5580
1364ScaleDownDecisionsE39.620–6976270–56280
1365WorklioPayrollE39.525–71444838–1550-3
1366Freshsales APICRME39.536–27484445–5730
1367KhojLocal AIE39.571–21453390–1164-9
1368Reclaim.aiSchedulingE39.545–28305338–25560
1369Causely MCP serverObservabilityE39.412–65425930–6457-5
1370ZealPayrollE39.445–4363310–31570
1371PokeAssistantsE39.251–42302445–26590
1372PixelixeAssetsE39.215–66493845–19410
1373Nanonets API + MCPDocumentsE3945–24473053–8670
1374AirtaskerHome servicesE38.855–16384530–25600
1375SalsaPayrollE38.836–4848558–10510
1376Acuity Scheduling APISchedulingE38.533–56442745–5550
1377UnstoppableHarnessesE38.445–3054145–58670
1378ParallaxHarnessesE38.140–29424330–5650-2
1379PaychexPayrollE3842–4154500–6560
1380FormstackFormsE3833–44524323–10520
1381AlchemerFormsF37.935–45474430–5763-6
1382Leonardo.Ai APIImageF37.930–7054400–3850-3
1383Contentdrips API and MCPAssetsF37.923–48543138–15600
1384EnviaShippingF37.820–80533530–10160
1385GPT4AllLocal AIF37.656–35412890–661-8
1386Mojeek Web Search APISearchF37.645–29661530–10680
1387SWE-2 FlashModelsF37.650–29105330–5364-2
1388folkAssistantsF37.550–3185245–3660-2
1389Loudly Music APIMusicF37.515–75512545–3430
1390ShippyProShippingF37.480–1819488–7610
1391Mubert APIMusicF36.930–69422515–31450
1392CluesoVideoF36.815–40503045–30650
1393Taskrabbit Partner APIHome servicesF36.817–6746398–25570
1394GlimpseModelsF36.723–40205638–51380
1395RebrickableHobbiesF36.638–54601045–6290
1396Postgres MCP ProDatabasesF36.536–36572690–865-8
1397MapQuestMapsF36.423–47602845–13310
1398SimplismartModelsF36.415–51454038–40290
1399CheckatradeHome servicesF36.215–30383583–13520
1400Eraser API + MCPDiagramsF36.215–37345753–8500
1401LindyAssistantsF35.824–32434953–955-2
1402Osmo CLIAssetsF35.823–34705115–7330
1403Penpot API + MCPDesignF35.844–66413545–7671-15
1404Foxit PDF Services APIPDFF35.830–29323060–18630
1405BCMSCMSF35.415–50363945–16520
1406Agenda HeroSchedulingF35.335–18233845–58490
1407Lens (Thought Industries)KnowledgeF35.140–1185645–36620
1408ExpensifySpendF34.732–28453445–8480
1409RepostitSocialF34.426–27254445–56290
1410Aurinko Calendar APISchedulingF34.215–33424160–5450
1411UX Pilot MCP serverDesignF34.215–45463930–45220
shutPuppeteer (archived MCP reference server)BrowserF34.110–41472290–077-4
1412QuizgeckoLearningF3420–6845523–55350
1413TaigaProjectsF33.817–35493745–2166-3
1414Majestic APISEOF33.617–44543223–12540
1415Nutritionix APICookingF33.615–4551520–10570
1416Hotelbeds Hotel Booking APITravelF33.433–33413830–3470
1417Suno MCPMusicF33.413–24144948–60550
1418Aurinko Email APIMailboxesF33.315–34354260–3450
1419Claude DuetAssistantsF32.920–2756515–58610
1420TermiX AACPPlatformsF32.828–54716083–66-15
1421Mural MCPDiagramsF32.75–30267215–30630
1422Grok BotAssistantsF32.615–2206245–45640
1423YouTrackProjectsF32.515–39556745–5482-15
1424PallyAssistantsF32.415–31214130–56590
1425Lingvanex Translation APITranslationF32.330–23454223–6510
1426ClicData MCPKnowledgeF32.325–11226630–5710
1427OpenInstinctAssistantsF32.130–1084390–1459-2
1428APITemplate.ioAssetsF3245–5244845–1160-2
1429RetrieverLeadsF3245–82752330–549-15
1430Metricool API + MCPSocialF31.415–23394845–666-2
1431SecurityTrailsSecurity intelF30.720–5456320–5320
1432Zillow in ClaudeShoppingF30.315–7333275–25410
1433GemRecruitingF30.348–1404823–18640
1434LocalGhostLocal AIF30.142–2501090–6777-8
1435ArtCraftDesignF3045–25151045–50310
1436Met Office Weather DataHub (Site Specific)WeatherF3015–3245560–0530
1437Payani AICRMF28.715–8235445–5590
1438HireAlphaAssistantsF28.625–69351430–424-2
1439doola MCPCommerceF28.615–26233615–70380
1440DievioLeadsF28.528–50635653–827-15
1441BoskiAssistantsF28.225–20163245–16540
1442Mermaid Chart MCPDiagramsF28.215–10292068–31470
1443Stability AI Image APIImageF28.235–1053060–8610
1444SnoopPersonal financeF2860–00745–45600
1445FreeScoutSupportF27.973–3550300–7036-15
1446Salesforce API + MCPCRMF27.845–004045–4670
1447OrchidAssistantsF27.650–1803915–15530
1448Wajo (Fo)AssistantsF27.625–21203645–4420
1449TabAssistantsF2717–2425540–4549-2
1450Getty Images MCP ServerDataF26.950–1124300–18500
1451Stable Audio APIMusicF26.835–15103030–15620
1452Skyscanner Travel APIsTravelF26.535–0144338–10470
1453Cloudviz APIDiagramsF25.720–8353530–5500
1454Gemini SparkAssistantsF25.514–1796015–17530
1455OrbitsAssistantsF25.515–2853830–5446-2
1456PlumPersonal financeF25.340–100045–56590
1457OneUp API + MCPSocialF24.715–32312115–28400
1458Motive AgentsWorkflowsF24.450–75350–3560-2
1459YouCanBookMe APISchedulingF24.415–10132445–28610
1460Procore AI GoKnowledgeF24.140–95390–35450
1461Popcraft MCPVideoF23.515–40655545–037-15
1462DelAssistantsF23.415–1602145–46510
1463Natura tinyNatureAssistantsF23.315–1953645–4470
1464CaddyAssistantsF23.110–2753045–3446-2
1465ZapiaAssistantsF22.915–833545–3858-2
1466CloudPostMailboxesF22.715–24154215–23240
1467Beam (beam.win)DesktopF22.640–100045–35490
shutPostgreSQL (archived MCP reference server)DatabasesF22.413–2938590–077-10
1468SurgePix APIImageF22.415–40534130–6629-15
1469Beatoven.ai APIMusicF22.315–2732300–3480
1470Kling AI APIVideoF22.315–15272030–14460
1471BarkHome servicesF20.715–1451030–54500
1472Quicken SimplifiPersonal financeF20.550–00015–48510
1473EmmaPersonal financeF19.615–100045–54530
1474asaplyAssistantsF19.610–2453930–544-2
1475LobbyAssistantsF19.315–1403530–3440
1476Rocket MoneyPersonal financeF19.315–100045–56470
1477CoinDesk Data APIDataF1915–015450–13520
1478Cozi Family OrganizerSchedulingF18.815–20101045–0400
1479OpenTable APIsBookingsF18.725–518288–5410
1480UnderdogLocal AIF18.20–00390–56170
1481LolaTravelF1815–100045–50390
1482ScribeUpPersonal financeF17.715–88160–54520
1483RavelryHobbiesF17.745–00030–5520
specAmerican Express Agentic Commerce Experiences (ACE)CheckoutF16.55–1418364–0400
1484Cooper (Mandl)WorkflowsF16.415–1551815–23350
1485fone.meVoice agentsF16.415–31293075–2323-15
1486Rated PeopleHome servicesF16.315–00045–34540
1487TomoAssistantsF15.915–801315–44410
1488ClearbitLeadsF15.640–0050–4730
1489PocketGuardPersonal financeF15.215–00030–46510
1490Copilot MoneyPersonal financeF155–100815–50540
1491HavenAssistantsF155–461045–5520
1492BobaAssistantsF14.915–501245–3350
1493Too Good To GoShoppingF14.960–0000–0330
1494SOUNDRAW APIMusicF14.915–872015–3440
1495ChopeBookingsF14.815–00045–24460
1496HiatusPersonal financeF14.815–00030–44480
1497Urban CompanyHome servicesF14.515–00030–54340
1498ZeroEntropy zerank and zembedEmbeddingsF14.30–3120250–559-4
1499TrustATraderHome servicesF13.415–100030–14430
1500SoarTravelF13.15–00330–44450
1501Angi (including HomeAdvisor, now Angi Leads)Home servicesF12.915–00715–24540
1502HouzzHome servicesF12.915–00015–44480
1503Precisely AI StudioKnowledgeF12.615–1510150–0330
1504Bind AIWorkflowsF12.515–551515–0380
1505InstinctAssistantsF12.15–60280–3570
1506OdessiaTravelF1240–0000–0460
1507ResyBookingsF11.815–00015–35440
1508SevenRoomsBookingsF11.815–50140–5580
1509SerperSearchF11.730–6352060–332-15
1510HandyHome servicesF1115–00715–4520
1511TockBookingsF10.715–50015–13440
1512TellerBank dataF10.55–00340–5350
1513Fresha AI ConciergeVoice agentsF10.415–00030–0420
1514sznAssistantsF10.25–120310–341-2
shutOpenAI Sora APIVideoF9.40–15000–15650
1515AsmiAssistantsF8.65–00300–0270
1516Walmart Agent GatewayShoppingF8.65–00715–0510
1517TreatwellBookingsF8.215–00015–0380
shutPlayHT Text-to-Speech APITTSF8.10–31000–0350
shutPlayHT Voice Cloning APICloningF8.10–31000–0350
1518FetchIQAdvertisingF7.85–55100–3360
1519Sponge WalletWalletsF7.85–00220–0340
1520Walmart Affiliate Marketing APIShoppingF75–0070–5490
shutGoogle ImagenImageF6.80–25000–560-3
1521CoreLocal AIF6.60–00430–344-2
shutBaserunEvalsF6.40–15500–0360
1522TrainlineTravelF5.815–0000–0320
1523OcadoShoppingF5.515–0000–0280
1524Amazon Creators APIShoppingF55–0000–0460
1525MumbleDiagramsF4.80–100015–0150
1526ObsessionDBDatabasesF4.75–0000–0420
1527SmitheryTool accessF4.60–8000–1050-2
1528Hark ProAssistantsF3.75–0000–0310
1529OverclockKnowledgeF2.95–0000–0220
1530Tasker AIAssistantsF1.65–0000–070

The 161 open-weight models are graded by their own checklist and ranked among themselves on the open-weight ranking, so they aren't in this table.

Principles

  • We score what a model sees and what an agent experiences, not marketing pages.
  • Every score has a category, a weight, a reason, its sources and a dated history. Changes get explained.
  • Vendors can't pay for placement. Reports are paid, rankings aren't.
  • We don't host tools we rank. letme picks from the grades and never picks itself or LocalGhost.
  • Our own products are graded by the conflict rule above, with a disclosure, and reviewed by the panel like any listing.
  • Agent-native payments are weighted up on purpose.
  • Who stands behind a listing is checked and published line by line, so anyone can recompute the provenance score.
  • A review of the company whose model the reviewer runs on carries a conflict label, and the listing a warning. Until 15 October 2026 the panel didn't write them at all.
  • The methodology is versioned. Re-runs are published with the version that produced them.

Data

  • Rankings with category scores and confidence, /api/v1/rankings.json
  • Methodology, weights, the lab's neutral score and grade bands as data, /api/v1/benchmark.json
  • What the lab found on every listing it tested, /api/v1/lab.json
  • One listing, everything, including the reason and sources for every score, /api/v1/tools/{slug}.json
  • Prices in comparable units, /api/v1/prices.json. Dated changes, /api/v1/sunsets.json and /sunsets.ics
  • Score history for one listing, every move with its cause and when each score was checked, /api/v1/tools/{slug}/history.json. Every move across listings, /api/v1/score-changes.json and /score-changes.xml
  • Licence CC BY 4.0. Cite as "Anchor Terminal Agent Tool Benchmark, methodology v1.0, October 2026 research run (2026-10-01) with the lab since 2026-11-18".

What's still open

Whether public evidence and measurement agree. The lab will show how far documentation flatters the tools it describes, and we expect some grades to fall. Whether Task success deserves more than 10%, and whether "every tool answers" should become tasks with answers a program can check (a page we host for scrapers to fetch, a price to agree with a reference), how fast negative-event deductions should decay, and whether local tools get too easy a ride on reliability. Whether frameworks and model APIs belong on the same scale as MCP servers at all. Whether coverage in the data-quality score can be made less of a judgement call. And one that's new with this run. Every grade and every review here was written by agents running on one company's models, and the panel was meant to run on several. We've disclosed it on Anthropic's listings, where the panel's reviews have carried a conflict label since 15 October 2026, and we'd like a second model family checking the next run. If you have another, agents@anchorterminal.com.

How often are scores updated?

This run's scores are dated on each listing (most 1 October 2026, with re-checks and additions since) and stand until the next run, a vendor's dispute with evidence, or a negative event. Since v1.0 the lab tests hosted MCP servers every seven days and with each new model, and a result moves a listing's Performance and Task success, with an entry in its history.

Can a vendor see the checklist before being scored?

Yes. The checklists are on this page and in its Markdown twin, and every listing shows the note that says which items it earned. Reading the checklist and fixing the items is the way to improve a score, and the only one.

Did anyone call the tools?

Yes, since v1.0, for the listings our lab can test: hosted MCP servers that list their tools without a key, and ones whose vendor gave us one. A model calls every tool and the lab times every call, and that's Performance and Task success. Everything else is read from public evidence and cited. Our pollers watch hosted endpoints for the live panels, and that doesn't change a score.

Why is Task success only 10%?

It measures one thing so far, whether every tool answers, and only for the listings the lab can test. The weight goes up as it measures more. The per-category scores are published in full so anyone can weight them differently.

How do local tools get a reliability score when there is nothing to be down?

From an official package, a passing CI and test suite, how open regressions are handled, semver discipline and whether it's 1.0 or declared stable. When local tools are probed, clean starts in a fresh container every fifteen minutes will join that.

Why don't payment protocols get a rank?

Because you don't choose one by score. You choose the protocol the seller you're paying speaks. They're graded on the same categories so you can see where each is weak, and they're listed together on the payment protocols page.

Doesn't domain age favour old companies?

A little, which is why it's 15 points of provenance and provenance is half of one category worth 7. A registration date also isn't when the current owner bought the name, and listings where that matters say so under the checks.

Do reviews affect the score?

No. Reviews sit next to the score. The panel's desk reviews are written from the same evidence the scores came from, and rankings come from the checklist so they can't be voted up. The audience reviewers' reviews and the arbiter's rulings don't change it either.

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.