Helicone AI Gateway + MCP by Helicone (Mintlify)

# Fix list: Helicone AI Gateway + MCP

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/helicone, the October 2026 research run, assessed 1 October 2026. Grade D, 47.1 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on Helicone AI Gateway + MCP: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reliability, 50 out of 100, up to 10 more on the total

Why it scored 50: Better Stack page at status.helicone.ai with three components (website, async logging endpoint, EU API) and uptime bars, but the AI Gateway agents are told to call isn't one of them (15 of 20). No incident write-ups at all. The bars show the async logging endpoint at 98.147 per cent and the EU API at 96.555 per cent, next to a note that every interruption in the last 90 days was a provider outage. The four-day logging and dashboard outage of 14 to 18 May 2026, which Helicone's own blog describes, appears nowhere on the page (5, no readable history). Ingestion and API rate limits published per plan on the pricing page (15). The gateway's error-handling page documents its 429s, but we found no Retry-After or backoff guidance for Helicone's own limits (5). No SLA found (0). Gateway and logging are GA (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

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.

## 2. Security & auth, 50 out of 100, up to 8.8 more on the total

Why it scored 50: Plain revocable Helicone API keys as a Bearer token. We found no documented scopes (20). No read-only switch on the MCP server, which mixes two reads with a tool that makes paid model calls (5). Logged prompts and responses come back to the agent as data, and we found no prompt-injection guidance for that path (5). Every gateway call is logged with its key, cost and properties, which is the product (12). The docs claim SOC 2. No security.txt, no SECURITY.md and no GitHub advisories, although two batches of authorisation fixes shipped on 30 August and 16 September and one commit mentions a disclosure programme that isn't published (8).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 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.

## 3. Payments & pricing, 30 out of 100, up to 8.8 more on the total

Why it scored 30: No x402 or other machine payment (0). Plan prices are public ($79 and $799 a month), but usage charges above 10,000 requests and 1 GB aren't itemised (10). Hobby is free with no card (20). A person has to sign up in a browser to get a key. Self-hosting is free but isn't an agent route (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

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

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. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 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 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 4. Schema & documentation, 69 out of 100, up to 5 more on the total

Why it scored 69: OpenAPI for the AI Gateway and a Swagger file for the REST API, both in the docs (25). llms.txt at docs.helicone.ai (10). The MCP tools have one-line descriptions that say what each does but not when to use it, and `use_ai_gateway` doesn't warn that it spends money (8). Filters are generated typed operators with enums, but `limit` has no bounds and `customProperties` is a free-form record (9). The gateway error-handling page lists codes and fixes, and most endpoint pages carry examples (12). The public changelog's newest entry is from 26 November 2025, and the 30 August API removals went out in commits and docs edits only (5).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

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.

## 5. Agent ergonomics, 70 out of 100, up to 4.9 more on the total

Why it scored 70: The MCP server has 3 tools (`query_requests`, `query_sessions`, `use_ai_gateway`), though the docs list 2. The recursive filter schemas add some weight (20). Offset, limit, filters, sort and `includeBodies` off by default (20). Gateway errors are documented with codes and fixes. The MCP server returns failures as plain text without the `isError` flag (12). No `readOnlyHint` or `destructiveHint` annotations and no idempotency keys, while one of three tools spends credit (3). Only time bounds are required. Helper SDKs for TypeScript and Python (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-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.

## 6. Maintenance & community, 66 out of 100, up to 3 more on the total

Why it scored 66: No tagged release since 2025-08-21 and no changelog entry since 26 November 2025, but the hosted service took deploys on 13 and 16 September 2026 carrying security fixes (30, counted as the last published API change). More than three deploy tags in 90 days, none with release notes (15). 56 open issues. We couldn't load reply times because GitHub's robots rules block issue searches. The product is in maintenance mode by its own notice (8). `@helicone/mcp` 0.1.6 was last published on 2025-11-04 (8). CI exists, and the 30 August commit says four workflows had been failing on main for lack of runner disk until then (5).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 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.

## 7. Transparency & trust, 71 out of 100, up to 2.5 more on the total

Made of editorial 60, provenance 82.

Why it scored 71: Apache-2.0 (30). The privacy policy was last updated on 28 February 2023, doesn't mention LLM request data, gives no retention period and links no DPA. Retention per plan is on the pricing page. A commit of 1 May 2026 says full request and response bodies had been written to Helicone's CloudWatch logs since 3 April, without a customer notice we could find (8). Dated notices for the Experiments removal and the Mintlify maintenance mode, and the removed Realtime page says so (12). The privacy policy names some processors (Cloudflare, PostHog, Stripe, Google Analytics) and says data is processed in the US. No full subprocessor list (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 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.

Provenance checks not met in full (half of this category, computed from checked facts):

- Domain age: helicone.ai, registered 2022-11-14 (3 years) (7 of 15)
- security.txt: not found (0 of 10)

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 2026-08-30. Helicone's own security commit (INC-657) fixed an authenticated cross-tenant read of other organisations' decrypted provider keys, a route that let any account mint proxy keys spending another organisation's provider key, a shared client that reused one organisation's OpenRouter key for others, an unauthenticated SSRF path and unauthenticated Stripe routes. Fixed, but disclosed only in the commit message, with no advisory or key-rotation notice that we found. -7 after decay for the fix (https://github.com/Helicone/helicone/commit/ca34549ea56f7ed587843f82d9cc19baa1f36ba4)
- 2026-09-16. A second fix closed a platform-admin takeover with a forged API key and a cross-tenant read through HQL, reported through Helicone's disclosure programme the same day. Fixed within hours and again disclosed only in the commit. -3 (https://github.com/Helicone/helicone/commit/067d9290acb4f1fc9320e902fc67b4b399b50363)

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- Whether the cross-tenant provider-key read fixed on 30 August 2026 was exploited, and whether affected customers were told, isn't stated anywhere we could find
- Whether organisations already using Helicone credits before 25 August 2026 were switched on automatically isn't documented
- Where the pricing page's 7-day trial buttons lead now that the in-app checkout is gone. We didn't sign up to test it
- GitHub issue reply times, which GitHub's robots rules kept us from loading

## Weaknesses

- Security fixes on 30 August and 16 September 2026 closed cross-tenant access to other customers' decrypted provider keys and an admin takeover, disclosed only in commit messages
- Maintenance mode since March 2026, changelog silent since 26 November 2025, and Experiments, Realtime and self-serve upgrades removed on 30 August 2026
- The status page has no incident history, leaves out the AI Gateway and doesn't mention the four-day outage of 14 to 18 May 2026
- Helicone credits need a per-organisation switch since 25 August 2026, so new accounts must bring their own provider keys
- Privacy policy last updated in February 2023, with no retention period or DPA

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Bring your own provider keys. Helicone credits return 403 unless support has enabled them for the organisation
- Rotate any provider keys stored in Helicone before 30 August 2026
- Leave `use_ai_gateway` out of the MCP client's allowed tools unless the agent is meant to spend on model calls
- Add `Helicone-Session-Id` and `Helicone-Session-Path` headers to group an agent's steps into one session
- Use the EU host (eu.helicone.ai) if the account was created there

## What the review panel asked for

- changelog entries for removals
- document `use_ai_gateway`
- add tool annotations

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.