Amplitude

# Fix list: Amplitude

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/amplitude, the October 2026 research run, assessed 7 October 2026. Grade B, 66.2 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 Amplitude: 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. Payments & pricing, 30 out of 100, up to 8.8 more on the total

Why it scored 30: Read with the hosted rubric. No x402, MPP or L402 (0). Plans are public but Plus shows a price only through an estimator that asks for a card, and Growth and Enterprise go through sales (10). The Free plan has 2 million events a month with no card, and MCP works on it (20). A person signs up and approves OAuth in a browser, or creates keys in settings (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.

## 2. Reliability, 60 out of 100, up to 8 more on the total

Why it scored 60: Read with the hosted lines and scored on what an agent calls, the remote MCP server first and the REST APIs beside it. status.amplitude.com on Statuspage has 29 components, MCP among them, with incident history (20). Between 9 July and 7 October 2026 it lists one critical and six major incidents. The critical one, on 13 July, affected the UI, MCP and REST APIs in the US from 14:04 PT and was resolved at 16:07 PT. The majors were on shared embeds, EU web latency, a metrics page, AI chat and two export delays on 22 and 23 September, so one outage reached the agent surface and the rest did not (5 of 30, between one major and several). Rate limits with numbers for the Dashboard REST API (5 concurrent, 108,000 cost an hour), Experiment Management (100 a second, 100,000 a day) and HTTP V2 (30 events a second per user or device), and none found for MCP or the Developer API (12 of 15). HTTP V2 says to pause 30 seconds on 429 and deduplicates on `insert_id`, and the Developer API returns `retryable` and `retry_after_seconds` and accepts `Idempotency-Key` on some operations. No guidance for MCP (13 of 15). No SLA found, and the terms disclaim uninterrupted service (0). The MCP server carries no beta label and is 1.0.0 in the official registry. The Developer API's spec is version 0.1.0 (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.

## 3. Schema & documentation, 72 out of 100, up to 4.6 more on the total

Why it scored 72: The Developer API has an OpenAPI 3.1.1 document whose root file is public, but its referenced path files returned 404 and the bundled /v1/openapi.json needs a token. The older REST APIs have no spec, and we couldn't read the MCP tool schemas without an Amplitude sign-in (15 of 25). llms.txt, five product feeds, llms-full.txt and Markdown for every docs page (10). The MCP page says what the server is for and what it isn't, and gives each of 45 tools a one-line description (15 of 20). REST parameter tables are typed with required fields, and the Developer API documents scopes, limits and formats. MCP input schemas unread (9 of 15). Request and response examples throughout the REST docs and error tables per API (13 of 15). Dated release notes at amplitude.com/releases and versioned API paths, but the registry entry has stayed at 1.0.0 since October 2025 while tools were consolidated on 17 August 2026 (10 of 15).

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.

## 4. Agent ergonomics, 72 out of 100, up to 4.6 more on the total

Why it scored 72: 45 tools is over the 30 line (5), with 10 added back for progressive discovery, which starts with four tools and loads schemas on demand, and a subset of about 25 for ChatGPT (15 of 25). Cursor and `limit` pagination on the Developer API (default 50, maximum 200), the Experiment Management API and the data warehouse tools, and filters on search and replay tools. The MCP page warns that hosts may truncate large charts (17 of 20). RFC 9457 errors with `error_code`, `retryable` and field errors on the Developer API and detailed 400 and 429 bodies on HTTP V2. MCP error shapes unread (15 of 20). `Idempotency-Key` and `dry_run` on some Developer API operations, `insert_id` on events, two-step deletes and optimistic concurrency on dashboards and notebooks. We couldn't see whether tools carry readOnlyHint or destructiveHint (13 of 20). Official SDKs in many languages cover ingestion and experiment evaluation, not queries, and the `amp` CLI prints compact JSON when piped (12 of 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.

## 5. Security & auth, 74 out of 100, up to 4.6 more on the total

Why it scored 74: The MCP server takes OAuth 2.0 with PKCE S256, dynamic client registration, a revocation endpoint and `mcp:read` and `mcp:write` scopes, and the Developer API takes scoped tokens. The older REST APIs still use a project API key and secret key, which aren't scoped (28 of 30). Two role actions, Use MCP (read) and Use MCP (write), are enforced per project on each call, an admin can switch MCP off for the organisation, deletes need a second confirmed call and cohort syncs preview first. MCP is on by default and three standard roles include write (17 of 20). Session replays, feedback comments and user properties are untrusted content, and we found only a disclaimer about model outputs, no injection guidance for hosts (2 of 15). An organisation audit log with 90 days of entries and an API, whose event list doesn't name MCP tool calls (9 of 15). security.txt valid to 31 December 2028, SOC 2 Type II, ISO 27001, 27017 and 27018 on the trust portal and a bug bounty that is private (18 of 20).

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.

## 6. Transparency & trust, 78 out of 100, up to 1.9 more on the total

Made of editorial 63, provenance 93.

Why it scored 78: Closed service with terms effective 21 July 2026. The SDKs, CLI and marketplace plugins are MIT (15). The DPA of 4 June 2024 promises destruction within 90 days of termination and use only to run the service, while the terms let Amplitude apply machine learning to customer data to improve the services and delete data at its discretion after termination. The AI docs say data isn't used for model training (20 of 30). A written SDK lifecycle with at least 12 months of maintenance for a superseded major version. No deprecation policy for the REST APIs or MCP tools found (10 of 20). Sub-processor list updated 13 May 2026 with US and EU data centre locations and 10 business days' notice of additions under the DPA (18 of 20).

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):

- Terms of service: read, states 6 of the 7 things a reader expects, and has 3 clauses that cost points (3.1 of 10)

## 7. Maintenance & community, 83 out of 100, up to 1.5 more on the total

Why it scored 83: @amplitude/developer-cli 0.4.0 on 2 October 2026, and the release notes add MCP taxonomy delete and restore on 12 September (30). At least 15 dated release notes since 17 August and four CLI releases since 5 August (20). Closed service with public release notes, a feedback form for MCP and 19 commits to the amplitude/mcp-marketplace repository since 9 July. We didn't test support (11 of 15). com.amplitude/mcp-server is in the official MCP registry under the vendor's domain (15). The marketplace plugin bumps versions automatically (1.6.0 on 8 September), while the CLI needs Node.js 26 and the registry entry is a year old (7 of 10).

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.

## 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.

- unchecked: the MCP tools' JSON Schema inputs and their readOnlyHint and destructiveHint annotations, which need an Amplitude sign-in to list
- unchecked: the Developer API's per-operation schemas, since the path files referenced by the public OpenAPI root returned 404 and /v1/openapi.json needs a token
- unchecked: the Plus plan's price per event, shown only in an estimator
- unchecked: whether MCP tool calls appear in the organisation audit log, which the audit log page doesn't say
- unchecked: how MCP service accounts are created, named on the MCP page with no setup guide found
- No rate limit for the MCP server or the Developer API was found in the reviewed documentation
- No SLA was found on amplitude.com (amplitude.com/sla returns 404). An enterprise contract may include one
- GitHub star counts weren't collected. npmWeekly is @amplitude/analytics-node and pypiWeekly is amplitude-analytics, both ingestion SDKs

## Weaknesses

- No rate limit for the MCP server or the Developer API was found in the reviewed documentation
- status.amplitude.com lists one critical and six major incidents between 9 July and 23 September 2026, with the UI, MCP and REST APIs down on 13 July
- No SLA found, and the terms of 21 July 2026 disclaim uninterrupted service
- MCP is on by default for every user, and the Member, Manager and Admin roles include write access until an admin changes them
- Session replays, feedback comments and user properties reach the model with no prompt-injection guidance for hosts

## 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.

- Connect to https://mcp.amplitude.com/mcp, or https://mcp.eu.amplitude.com/mcp for EU projects. The client must support streaming HTTP and OAuth, since API keys aren't accepted
- Append `?discovery=progressive` to load tool schemas on demand, then call `get_amplitude_context` first to pick the project and read its AI context
- Ask an admin for a role with only Use MCP (read) on the projects you need. Tools stay listed and fail at call time when the role lacks the action
- Send events through the HTTP V2 API with an `insert_id` on each one. On 429, pause that user or device for 30 seconds before retrying
- Budget Dashboard REST queries by cost (days times conditions times chart cost), within 108,000 an hour and 5 concurrent requests per project

## 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.