Amplitude MCP Server

Official Amplitude MCP server with 24+ tools for dashboards, charts, experiments, custom queries, session replays, feature flags, and event properties.

Analytics by Amplitude OAuth2 active

Connect (Remote Server) Docs

Overview

The Amplitude MCP server is the official, provider-hosted Model Context Protocol endpoint maintained by Amplitude. It lets AI assistants query and modify Amplitude analytics content (dashboards, charts, notebooks, experiments, feature flags, cohorts, session replays, agent analytics, and feedback) using natural language. Authentication is handled via OAuth 2.0 with your Amplitude account, so the server respects your existing organization, project, and permission scopes: you can only access data you can already see in the Amplitude UI.

The server exposes a discovery layer (get_context, list_tool_categories, get_category_tools, describe_tool) and a "progressive discovery" mode that limits initial tool exposure to reduce token spend. Tool categories include core analytics (search, query_chart, query_experiment, query_dataset, query_metric), content management (get/create/edit dashboards and notebooks, get_charts, get_experiments), feature flags (get_flags, create_flags, update_flag), session replay (get_session_replays, list_session_replays, get_session_replay_events), AI agent analytics (query_agent_analytics_metrics/sessions/spans, conversation tools), and AI feedback (get_feedback_trends).

It is available to all Amplitude customers including the free Starter plan, with US and EU residency endpoints. Native connectors exist for Claude, Cursor, ChatGPT, Claude Code, Gemini CLI, Codex CLI, Figma Make, Kiro, Lovable, Replit, Notion, and others. Amplitude also publishes a companion mcp-marketplace repo of skills and agent prompts for use with the server.

Tools

Tool	Description
`get_context`	Confirm organization and project access; returns user and org context.
`list_tool_categories`	List available tool categories (analytics, dashboards, experiments, etc.) for progressive discovery.
`get_category_tools`	List tools available within a specific category.
`describe_tool`	Get the latest schema and required parameters for a specific tool.
`search`	Find dashboards, charts, notebooks, experiments, events, properties, and cohorts.
`query_chart`	Retrieve data from a specific chart.
`query_experiment`	Analyze experiment results and statistical significance.
`query_dataset`	Execute custom analytics queries against event data, including Network Request, Status Code, and Error Logged data.
`query_metric`	Get data for a specific metric.
`get_charts`	Retrieve full chart definitions by ID, including events, properties, and configuration.
`get_dashboard`	Retrieve a dashboard with all its charts and layout.
`get_notebook`	Fetch notebook content including embedded charts and analysis.
`get_experiments`	Retrieve detailed experiment information.
`create_chart`	Create a new chart in Amplitude.
`create_dashboard`	Create a new dashboard.
`edit_dashboard`	Modify an existing dashboard.
`create_notebook`	Create a new notebook.
`edit_notebook`	Edit an existing notebook.
`get_flags`	Retrieve feature flag configurations.
`create_flags`	Create new feature flags.
`update_flag`	Update an existing feature flag.
`get_session_replays`	Search and filter session recordings by events or user properties.
`list_session_replays`	List session replays for a project via the Session Replay public API.
`get_session_replay_events`	Retrieve events for a specific session replay.
`get_event_properties`	Explore event property definitions in the project taxonomy.
`query_agent_analytics_metrics`	Query metrics for AI agent analytics.
`query_agent_analytics_sessions`	Query AI agent sessions.
`query_agent_analytics_spans`	Query AI agent spans for tracing.
`get_agent_analytics_conversation`	Retrieve a specific AI agent conversation.
`search_agent_analytics_conversations`	Search AI agent conversations.
`get_agent_analytics_schema`	Get the schema for AI agent analytics data.
`get_feedback_trends`	Surface trends from AI feedback data.

Setup Guide

Prerequisites

An Amplitude account (any tier, including free Starter)
An MCP-compatible client (Claude Desktop, Claude Code, Cursor, ChatGPT, Gemini CLI, Codex CLI, etc.)
Permissions in Amplitude for the projects you want to query

Server endpoints

US (default): https://mcp.amplitude.com/mcp
EU residency: https://mcp.eu.amplitude.com/mcp
Progressive discovery (lower token cost): append ?discovery=progressive

Authentication

OAuth 2.0 with Amplitude. On first connection your MCP client will open a browser window to log in. No API keys needed. Access is scoped to your existing Amplitude permissions.

Claude Desktop / Cursor config

Add the remote server to your client's MCP config. For Claude Desktop, edit claude_desktop_config.json:

{
  "mcpServers": {
    "amplitude": {
      "url": "https://mcp.amplitude.com/mcp"
    }
  }
}

For EU data residency, swap the URL for https://mcp.eu.amplitude.com/mcp. To use progressive discovery, append ?discovery=progressive.

Claude Code

claude mcp add --transport http amplitude https://mcp.amplitude.com/mcp

Then run /mcp inside Claude Code to complete OAuth.

Notes

The server is in active development; new tools are added regularly (Session Replay, Agent Analytics, and Feature Flag tools are recent additions).
Large chart payloads may be truncated by some AI clients.
Admins can restrict MCP access org-wide via Amplitude settings.

Use Cases

Pull weekly retention, conversion, and engagement metrics into a Claude or Cursor chat without writing chart queries by hand.
Analyze A/B test results: ask the agent to summarize statistical significance and lift across variants for a running experiment.
Triage user issues by searching session replays filtered by error events, then inspecting the replay event stream.
Create or edit dashboards and notebooks from natural language ("build a dashboard tracking signup-to-activation by acquisition channel").
Manage feature flags from your IDE: list flags, create new ones, or toggle rollout percentages alongside the code you're shipping.
Investigate AI agent performance using Agent Analytics tools to query spans, sessions, and conversation traces.

Example Prompts

"Find the top 10 events driving drop-off between signup and first activation in the last 30 days."
"Summarize the results of the checkout-redesign-v2 experiment and tell me if any variant is significant."
"Show me session replays from yesterday where users hit a 500 error on the checkout page."
"Create a dashboard called Weekly Growth with charts for new signups, DAU, and 7-day retention."
"List all feature flags with rollout above 50% and tell me which are still tied to active experiments."
"Query agent analytics for conversations where the AI returned a tool error in the last 24 hours."

Pros

Official, Amplitude-maintained server with OAuth 2.0 and existing permission enforcement, no API key management.
Broad coverage: 24+ tools spanning analytics, dashboards, experiments, feature flags, session replay, agent analytics, and feedback.
Available on the free Starter plan with both US and EU residency endpoints.
Progressive discovery mode reduces upfront token cost for large tool catalogs.

Limitations

Provider-hosted only; you cannot self-host or fork the server.
Large chart or dataset responses may be truncated by some AI clients.
Tool surface is still expanding and currently in beta, so behavior and tool names can change.

Alternatives

PostHog MCP server for product analytics with a similar feature set on an open-source stack.
Mixpanel integrations via community MCP servers for event analytics.
Google Analytics MCP servers (community) for web analytics use cases.