Amplitude runs an official MCP server with both US and EU endpoints. It covers the questions product teams actually ask — funnel drop-off, retention by cohort, segment comparisons — without anyone opening the chart builder.
https://mcp.amplitude.com/mcp
Claude Sonnet 4.5
Chat with 60+ AI models on the same workflow — switch to a different model mid-conversation and re-run the same prompt, or use Compare mode to put several side-by-side and balance quality vs. cost.
Amplitude API credentials via an authorization header. EU projects must use https://mcp.eu.amplitude.com/mcp instead of the US endpoint.
How models use it and what it is built for.
The server exposes Amplitude’s analytics surface as tools an agent can compose. Event queries are the base: count and segment events over a time range, filtered by user or event properties. On top of that sit the analyses Amplitude is known for — funnels with conversion and drop-off per step, retention curves by cohort, and segmentation that compares one user group against another. A model can also inspect the event taxonomy, which is the unglamorous but decisive part: knowing which events actually exist, and what properties they carry, is what separates a real answer from a confident guess about an event name that was never instrumented. Amplitude operates separate US and EU deployments and the endpoints are distinct, so EU customers connect to mcp.eu.amplitude.com and their data stays in the region.
Typical tools an AI model can call. Exact names vary by version.
Taken from the official Amplitude documentation — see Amplitude documentation for the full reference.
Examples
US
https://mcp.amplitude.com/mcpEU
https://mcp.eu.amplitude.com/mcpRequired for EU-resident projects — the US endpoint cannot see EU data.
Copy any of these into MCP Agent Studio after connecting.
Where do users drop off between signup and first project created?
Compare week-four retention for users who came from paid versus organic.
Which events fired most often last month, and which have stopped firing entirely?
Has activation improved since the onboarding change on the first of the month?
This is not a single-model product: you get the same MCP connection with 60+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.
Default pick for Amplitude
Claude Sonnet 4.5
Funnel and retention definitions are easy to get subtly wrong — wrong start event, wrong window. Sonnet 4.5 states its assumptions, which is what makes the number checkable.
Listing tools proves the server is reachable, not that a model can work with it. Evals go further: they read every tool on the server, write a test suite from its real schemas, and run it — code decides pass/fail on the responses (schema conformance, error codes, pagination, result caps) while a scoring model grades plain-English tasks driven through the tools.
Get a pass/fail report per tool with the evidence behind each verdict — and replay the same suite after every schema change. Destructive tools are excluded from the run.
Run evalsOpen MCP Agent Studio with the connection pre-filled. Add your token, pick any of 60+ models, and start chatting — no install required.
Open Agent StudioCommon questions about connecting, scoping and using it safely.
Amplitude’s official hosted MCP server. It exposes event segmentation, funnels, retention, cohorts and the event taxonomy as MCP tools, so an AI assistant can answer product analytics questions directly against your data.
Yes. EU-resident projects use mcp.eu.amplitude.com rather than mcp.amplitude.com. They are separate deployments — pointing an EU project at the US endpoint will not find your data.
No, and this is the most common source of confusion. The server can only analyse what your SDK actually sends. Ask it to list the event taxonomy first — the answer to "why is this funnel empty" is usually a missing event, not a broken query.
Amplitude leans on named analyses — funnel, retention, segmentation — as distinct tools. PostHog leans on HogQL, a SQL dialect, so it is more flexible but demands more query correctness. Amplitude is easier to get a right answer from; PostHog can answer stranger questions.
It is lower risk than most write operations, since a cohort is an analysis artefact rather than production config. The thing to watch is clutter — an agent that creates a cohort per question leaves a workspace nobody can navigate.
PostHog
Ask product analytics questions and ship feature flags from chat.
ClickHouse
Run analytical SQL over billions of rows in natural language.
Snowflake
Query the warehouse through Cortex Analyst, Search and Agents.
Datadog
Query metrics, logs, traces and monitors from a conversation instead of a dashboard.