Best of · Work & business apps
Best product analytics and experimentation tools for AI agents
The 10 highest-scoring of 13 product analytics and experimentation tools on the Anchor benchmark, with a pick for each need and where each one falls short. Scores come from public evidence, re-checked as vendors change.
- 13 ranked
- 2 agent-ready
- 12 hosted endpoints
- Updated 9 October 2026
Top three
Picks by need
Worked out from the scores, prices and facts, so they change when the research does.
Schema & documentation
89/100 on schema & documentation, against 82 for the overall leader.
Maintenance & community
GrowthBook BB
89/100 on maintenance & community, against 81 for the overall leader.
Self-hosting under an open licence
GrowthBook BB
self-hosted, MIT for the core and for the MCP server licence.
Also Matomo, self-hosted, Matomo core is GPL-3 licence.
The shortlist
| # | Tool | Grade | Best for | Price | Where |
|---|---|---|---|---|---|
| 1 | Statsig Amplitude, Inc. |
BB 72.3 | An agent that creates, updates and cleans up feature gates, dynamic configs and experiments, and reads experiment results and audit history. | $150 / mo | hosted |
| 2 | GrowthBook GrowthBook, Inc. |
BB 70.1 | Teams that want flags and experiments analysed against their own warehouse, with an open-source core and a self-hosted option. | $40 / seat-mo | hosted and local |
| 3 | LaunchDarkly Catamorphic, Co. (dba LaunchDarkly) |
B 69.2 | An agent that creates and targets feature flags, runs rollouts and experiments, reads experiment results and cleans up stale flags. | $10 / mo | hosted and local |
| 4 | PostHog PostHog Inc. |
B 68.4 | An agent that answers product questions in SQL or with funnel, retention and trends queries, and that manages feature flags, experiments and error tracking in the same project. | $0.0001 / tx | hosted |
| 5 | Amplitude Amplitude, Inc. |
B 66.2 | A team already on Amplitude that wants an agent to query charts, funnels, retention and experiment results, edit dashboards and manage the tracking plan under its own permissions. | Freemium | hosted |
| 6 | Fullstory Fullstory, Inc. |
B 62.6 | An agent that sends server-side events and user properties, fetches a session's events or an AI summary for support or debugging, or exports a segment. | Freemium | hosted |
| 7 | Optimizely Experimentation Optimizely |
B 62.6 | An agent working for an existing Optimizely customer that lists and creates flags, experiments and audiences, reads experiment results and writes SDK integration code. | Paid | hosted |
| 8 | Mixpanel Mixpanel, Inc. |
B 62 | Teams already tracking in Mixpanel who want an agent to answer funnel, retention and cohort questions or manage Lexicon, experiments and flags. | $140 / mo | hosted |
| 9 | Matomo InnoCraft Limited |
C 59.5 | Teams that want web and product analytics with data kept in the EU or on their own servers, queried by report, segment, funnel or cohort. | $26 / mo | local |
| 10 | Countly Countly Ltd |
C 54.8 | An agent whose owner runs Countly on its own infrastructure or a dedicated Flex server and wants analytics, crash and remote-config data through MCP with per-category permissions. | $175 / mo | hosted and local |
3 more are ranked in the full table.
How to choose
- Answers against known valuesAsk a funnel and a retention question whose answer you already know, since a query that returns a plausible number can still count events wrongly.
- Query limits and result sizeCheck the limits on query rate, result size and time range, since an agent that loops over a large cohort can hit them partway through an analysis.
- Raw event export and retentionCheck whether raw events can be exported and how long they are kept, since an agent cannot re-run a question once the events have aged out.
- Experiment results over the APICheck whether experiment results and flag states can be read through the API, since an agent deciding a rollout needs the same numbers the dashboard shows.
How the benchmark tests this category. One fixed event set loaded into each listing, then the same questions asked through its API or MCP server (a funnel conversion, a retention curve, a cohort size and an experiment result). We check the answers against known values and the limits on queries. In this run listings are graded from public evidence against the published checklist.
Each one in detail
Statsig
BB 72.3/100Statsig is a hosted platform for feature flags, experiments and product analytics, run by Amplitude, Inc. since May 2026. Agents reach it through the Console API, which has a public OpenAPI spec, and an official hosted MCP server.
Verdict Statsig suits agents that manage feature gates and experiments and read their results. Its MCP server has 18 default tools, OAuth, read-only access by project and role, and a confirmation on updates. Errors carry only a status and a message, no idempotency keys were found, and the privacy notice is Amplitude's and doesn't name Statsig.
Choose it for An agent that creates, updates and cleans up feature gates, dynamic configs and experiments, and reads experiment results and audit history.
Strengths
- Hosted MCP server with 18 default tools and a discovery layer that replaced 93 tools on the V1 server
- MCP access set per project and per role as no access, read only or read and write, with a confirmation on updates
- Public OpenAPI 3.0 spec with 325 operations, plus llms.txt and a Markdown copy of every docs page
Weaknesses
- Console API errors carry only
statusandmessage, and the spec documents no 429 response - No idempotency keys were found, and most MCP update tools replace the whole resource
- The project Console API key reads and writes everything in a project unless created read-only
Price $150 / moAuth OAuth or keyx402 nohosted
GrowthBook
BB 70.1/100GrowthBook is an open-source platform for feature flags, experiments and product analytics, sold as GrowthBook Cloud or run self-hosted. Agents reach it through a REST API with a public OpenAPI spec and an official MCP server.
Verdict GrowthBook Cloud's MCP server has four tools with read and write annotations, OAuth with dynamic client registration, and a 414-operation OpenAPI spec behind it. The free plan needs no card. The API tools pass paths and JSON bodies through without validation, every key is limited to 60 requests a minute, and audit logs are Enterprise only.
Choose it for Teams that want flags and experiments analysed against their own warehouse, with an open-source core and a self-hosted option.
Strengths
- The MCP server has four tools, each with readOnlyHint and destructiveHint, and a second endpoint with only the two API tools
- OAuth with PKCE, dynamic client registration, refresh and revocation on GrowthBook Cloud, so no API key sits in the MCP config
- Public OpenAPI 3.1 spec with 306 paths and 414 operations, plus llms.txt and a Markdown copy of every docs page
Weaknesses
- The API tools take a path and a JSON string and don't validate payloads, so correct calls depend on the bundled skills
- API keys are limited to 60 requests a minute on Cloud, and no idempotency keys were found
- Audit logging is an Enterprise feature, and Free plan members can only hold the Admin role
Price $40 / seat-moAuth OAuth or keyx402 nohosted and local
LaunchDarkly
B 69.2/100LaunchDarkly is a hosted feature flag and experimentation platform from Catamorphic, Co., with observability and AI configuration products. Agents reach it through an official hosted MCP server with OAuth and a REST API with a public OpenAPI spec.
Verdict LaunchDarkly suits agents that manage feature flags, rollouts and experiments. The hosted MCP server uses OAuth with dynamic client registration and marks its tools read-only or destructive, and the Developer plan is free without a card. The server lists 141 tools, numeric rate limits are unpublished, and the web application was unavailable for over an hour on 10 July 2026.
Choose it for An agent that creates and targets feature flags, runs rollouts and experiments, reads experiment results and cleans up stale flags.
Strengths
- Public OpenAPI 3.0.3 spec with 401 operations, 399 of them described, with 1,853 examples and 429 documented on 287 operations
- Hosted MCP server with OAuth, PKCE, dynamic client registration, a revocation endpoint and reader, writer and observability scopes
- MCP tools cover experiment creation, iteration control and results, flag targeting, approval requests and flag removal readiness
Weaknesses
- The hosted MCP server lists 141 tools, and no tool filter for it was found in the reviewed documentation
- The API docs say the numeric rate limits are not published and may change
- 132 of 401 API operations are beta, need
LD-API-Version: betaand may change without notice
Price $10 / moAuth OAuth or keyx402 nohosted and local
PostHog
B 68.4/100PostHog is an open-source product analytics platform with session replay, feature flags, experiments, error tracking and a data warehouse. Agents reach PostHog Cloud through a REST API with a public OpenAPI spec and an official hosted MCP server.
Verdict PostHog suits agents that need to query product data and manage flags or experiments. Its hosted MCP server has OAuth scopes, a read-only mode and a one-tool CLI mode over 1,096 tools. The status page shows 31 incidents in 90 days, four on analytics queries, and three security incidents were disclosed in the last year.
Choose it for An agent that answers product questions in SQL or with funnel, retention and trends queries, and that manages feature flags, experiments and error tracking in the same project.
Strengths
- Public OpenAPI 3.1 spec with 2,347 operations and a scope on most of them, plus llms.txt and a Markdown copy of every docs page
- Hosted MCP server with OAuth, a read-only switch, filters by product area or tool name and pinning to one project
- CLI mode registers one
exectool that searches, inspects and calls the 1,096 tools on demand
Weaknesses
- 31 incidents on the status page between 9 July and 6 October 2026, four of them analytics query timeouts or failures
- Three security incidents in twelve months, among them malicious npm SDK versions on 24 November 2025
- API queries stop at 10 seconds of execution, three at a time per project, with an hourly read budget on personal API keys
Price $0.0001 / txAuth OAuth or keyx402 nohosted
Amplitude
B 66.2/100Amplitude is a hosted product analytics platform that records user events and answers questions about funnels, retention, cohorts, experiments and feature flags. Agents reach it through an official remote MCP server, REST APIs and the amp command line.
Verdict The remote MCP server covers queries, funnels, retention, experiments and flags under OAuth with PKCE, read and write scopes and per-project role actions, and the Free plan needs no card. No MCP rate limits or SLA were found, and status.amplitude.com records a critical outage of the UI, MCP and REST APIs on 13 July 2026.
Choose it for A team already on Amplitude that wants an agent to query charts, funnels, retention and experiment results, edit dashboards and manage the tracking plan under its own permissions.
Strengths
- Remote MCP server with 45 documented tools for queries, funnels, retention, cohorts, experiments, flags and the tracking plan, on US and EU hosts
- OAuth with PKCE, dynamic client registration, a revocation endpoint and
mcp:readandmcp:writescopes, plus two role actions enforced per project on every call - Progressive discovery (
?discovery=progressive) starts with a small tool list and loads schemas on demand
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
Price FreemiumAuth OAuth or keyx402 nohosted
Fullstory
B 62.6/100Fullstory records web and mobile sessions and turns them into behavioural analytics. Agents reach it through a Server API for events, users, sessions and exports, and through a hosted MCP server, in beta, for metrics, funnels and journeys.
Verdict The Server API suits an agent that sends events or pulls session context. Create calls take an idempotency key and errors carry stable codes. Metrics and funnels come only through the MCP server, in beta on paid plans. Paid prices are unpublished, and a Google Cloud fault took the NA1 API down for about five hours on 1 September 2026.
Choose it for An agent that sends server-side events and user properties, fetches a session's events or an AI summary for support or debugging, or exports a segment.
Strengths
- Every create call accepts an
Idempotency-Keyheader, kept for 24 hours and checked against the original request body - API keys have three permission levels, and a Standard key cannot export or delete user data
- The MCP server takes OAuth with PKCE, 16 scopes and dynamic client registration at
https://api.fullstory.com/mcp/fullstory
Weaknesses
- Metrics, funnels and journeys are reachable only through the MCP server, which is in beta, limited to paid plans and outside the support SLAs
- No price is published for Business, Advanced or Enterprise. The plans page asks for a demo
- A Google Cloud fault took multiple NA1 services, the API included, out for about five hours on 1 September 2026
Price FreemiumAuth OAuth or keyx402 nohosted
Optimizely Experimentation
B 62.6/100Optimizely Experimentation is a hosted platform for A/B tests, feature flags and personalisation, covering Web Experimentation and Feature Experimentation. Agents reach it through a hosted MCP server with seven tools and through REST APIs.
Verdict Optimizely Experimentation suits agents working for an existing customer that manage flags and experiments and read results. The hosted MCP server has seven tools behind OAuth and carries the user's permissions, and the main REST API has a public Swagger spec. No price, free plan or trial is published, and no 429 or idempotency guidance was found.
Choose it for An agent working for an existing Optimizely customer that lists and creates flags, experiments and audiences, reads experiment results and writes SDK integration code.
Strengths
- Hosted MCP server with seven tools for querying, managing and implementing flags and experiments, released on 29 April 2026
- MCP sign-in is OAuth 2.0 with PKCE and dynamic client registration, and the server sees only what the user's account can see
- Public Swagger 2.0 spec for the Optimizely API with 93 operations, each with a description, plus llms.txt and a Markdown copy of every docs page
Weaknesses
- No price is published. The plans page says every plan is individually packaged and leads to a demo request
- No free plan or trial was found, and the MCP server needs an account with Opal enabled
- No 429 response,
Retry-Afterheader or idempotency key was found in the spec or the pages read
Price PaidAuth OAuth or keyx402 nohosted
Mixpanel
B 62/100Mixpanel is a hosted product analytics service that records events and answers questions about funnels, retention, cohorts, experiments and feature flags. Agents reach it through REST APIs with public OpenAPI specs and a hosted MCP server.
Verdict Mixpanel publishes 14 OpenAPI specs, llms.txt and a hosted MCP server with OAuth scopes, and a free plan covers 1M events a month without a card. The Query API is limited to 60 queries an hour and five at once, and the MCP docs describe no confirmation step for tools that delete or bulk edit.
Choose it for Teams already tracking in Mixpanel who want an agent to answer funnel, retention and cohort questions or manage Lexicon, experiments and flags.
Strengths
- 14 public OpenAPI specs covering 106 operations, plus llms.txt, llms-full.txt and a Markdown copy of every docs page
- Hosted MCP server in US, EU and India regions with OAuth, PKCE, dynamic client registration and 24 scopes
- Free plan with 1M events a month and no card, and MCP on by default for free and Growth accounts created after 1 August 2026
Weaknesses
- Query API limit of 60 queries an hour and five concurrent per project, and 600 MCP requests an hour per user
- 64 tools named on the MCP page, with deletes and bulk edits and no confirmation step described
- Query API returned HTTP 500 for US projects for about 77 minutes on 26 August 2026, per Mixpanel's post-mortem
Price $140 / moAuth OAuth or keyx402 nohosted
Matomo
C 59.5/100Matomo is an open-source web and product analytics platform from InnoCraft, sold as Matomo Cloud or run on the owner's servers. Agents query reports and manage configuration through its Reporting HTTP API or the official MCP server plugin.
Verdict The Reporting API covers reports, funnels, cohorts and experiments with an OpenAPI 3.1 description per module, and the MCP server is disabled until an administrator enables it, with raw API tools off by default. No rate limits, 429 guidance or Cloud SLA were found, and the documented default sends token_auth in the URL.
Choose it for Teams that want web and product analytics with data kept in the EU or on their own servers, queried by report, segment, funnel or cohort.
Strengths
- The API reference lists 72 modules, each with an OpenAPI 3.1 description embedded in its page, including Funnels (17 operations), Cohorts and A/B testing.
- The official MCP server plugin is included with Matomo Cloud, has 19 typed tools with read-only, destructive and idempotent annotations, and hides its seven raw API tools by default.
- An OAuth 2.0 authorisation server plugin issues bearer tokens with one of four scopes, with PKCE, client credentials and secret rotation.
Weaknesses
- No API rate limit numbers, 429 or retry guidance, or idempotency keys were found. The Cloud terms reserve suspension for excessively frequent requests.
- The API docs introduce authentication as a
token_authURL parameter. POST-only tokens and theAuthorizationheader are the recommended alternatives. - No SLA for Matomo Cloud was found. The terms supply the service as is, with email support on a reasonable effort basis.
Price $26 / moAuth OAuth or keyx402 nolocal
Countly
C 54.8/100Countly is a product analytics platform sold as a managed private-cloud server (Flex), a self-hosted Enterprise edition and the open-source Countly Lite. Agents reach a Countly server through its HTTP API or the official countly-mcp-server.
Verdict Countly suits an agent working against a server its owner already runs. The official MCP server filters its 209 tools by edition, plugin and permission and has per-category read-only settings. No status page, OpenAPI document or published API rate limit was found, and the Flex trial needs a credit card.
Choose it for An agent whose owner runs Countly on its own infrastructure or a dedicated Flex server and wants analytics, crash and remote-config data through MCP with per-category permissions.
Strengths
- Official MIT MCP server,
countly-mcp-server1.7.0 of 7 October 2026, with read-only, destructive and idempotent annotations on its tools COUNTLY_TOOLS_ALL=Rmakes every MCP tool read-only, and per-category settings take None, R, CR, CRU or CRUD- Auth tokens can be limited to named apps and endpoint patterns, carry a lifetime in seconds and are revocable
Weaknesses
- No public status page or incident history was found for Flex servers or for mcp.count.ly
- No OpenAPI document was found, and the server's API rate limit is a setting that ships switched off with no published value for Flex
- The API accepts
api_keyandauth_tokenin the URL query string, and theapi_keyhas its user's full read and write access
Price $175 / moAuth API keyx402 nohosted and local
Head to head
- GrowthBook vs Statsig BB 70.1 vs BB 72.3
- LaunchDarkly vs Statsig B 69.2 vs BB 72.3
- PostHog vs Statsig B 68.4 vs BB 72.3
- Amplitude vs Statsig B 66.2 vs BB 72.3
- GrowthBook vs LaunchDarkly BB 70.1 vs B 69.2
- GrowthBook vs PostHog BB 70.1 vs B 68.4
- Amplitude vs GrowthBook B 66.2 vs BB 70.1
- LaunchDarkly vs PostHog B 69.2 vs B 68.4
- Amplitude vs LaunchDarkly B 66.2 vs B 69.2
- Amplitude vs PostHog B 66.2 vs B 68.4
Questions
What are the highest-rated product analytics and experimentation tools for AI agents?
Statsig has the highest benchmark score of the 13 ranked product analytics and experimentation tools, 72.3 (BB). GrowthBook is second with 70.1 (BB).
How many product analytics and experimentation tools are agent-ready?
2 of the 13 ranked here grade BB or better, the bar for agent-ready on the Anchor benchmark.
Which product analytics and experimentation tools accept x402 payments?
None of the ranked listings here accepts x402 for its main call yet.
How is this list ranked?
By the Anchor benchmark score out of 100, a weighted mean of the scored categories minus deductions for negative events, from public evidence re-checked as vendors change. Listings cannot pay for a place. The latest assessment behind this page is from 9 October 2026.
How this list is made
The order is the Anchor benchmark score, the same number as on each listing and in the top list. Each listing is graded from public evidence against the benchmark checklist, and the picks above are worked out from those grades, prices and facts. No listing pays for its place, and paid audits or listing help never change a score.
Full ranked table · 73 head-to-head comparisons · Best tools in every category