Skip to main content
hc
Submit
mcpSource-backedReview first Safety ✓ Privacy ✓

Amplitude MCP Server for Claude

Official remote MCP server for querying and creating Amplitude analytics, dashboards, experiments, cohorts, session replay insights, and product data from Claude and other MCP clients.

by Amplitude·added 2026-06-05·
Claude CodeClaude Desktop
HarnessClaude CodeClaude Desktop
Review first — review before installing

Open the source and read safety notes before installing.

Safety notes

  • The authenticated Amplitude MCP server can expose tools that create and edit analytics content, not just read existing dashboards.
  • Use the existing Amplitude permission model and keep users scoped to the projects and organizations they are allowed to inspect.
  • Require human confirmation before Claude creates dashboards, edits notebooks, changes experiments, creates cohorts, updates metrics, or touches feature-flag workflows.
  • Use the EU MCP endpoint only when the Amplitude data resides in the EU region; otherwise use the standard US endpoint.
  • Consider progressive discovery to reduce tool-list size and force the assistant to describe a tool schema before calling higher-impact tools.
  • Organization administrators can restrict or disable MCP access in Amplitude content access settings; use that control if AI access is not approved.

Privacy notes

  • Amplitude MCP can expose product analytics, event data, user cohorts, experiment results, dashboard definitions, session replay context, feedback summaries, AI agent analytics, and account health signals to the connected AI client.
  • AI providers may process the Amplitude data returned through MCP; review internal AI, privacy, GDPR, CCPA, and customer-data policies before connecting production projects.
  • Assistant transcripts, MCP client logs, screenshots, support tickets, and exported chats can retain user identifiers, cohort definitions, behavioral data, experiment details, and roadmap-sensitive product insights.
  • OAuth ties access to the current Amplitude user, but it does not make downstream model transcripts private by itself.

Prerequisites

  • Existing Amplitude customer account with access to the projects you want Claude to analyze.
  • MCP-capable client that supports remote Streamable HTTP servers, such as Claude Code, Cursor, ChatGPT, Codex CLI, Lovable, Kiro, or another compatible client.
  • OAuth authorization through Amplitude for the user and organization being queried.
  • Organization-level MCP access left enabled by an Amplitude administrator.
  • Clear policy for whether AI tools may create or edit Amplitude charts, dashboards, notebooks, experiments, cohorts, metrics, feature flags, or annotations.

Schema details

Install type
cli
Troubleshooting
No
Source repository stats
Scope
Source repo
Collection metadata
Estimated setup
10 minutes
Difficulty
intermediate
Tool listing metadata
Full copyable content
{
  "mcpServers": {
    "amplitude": {
      "url": "https://mcp.amplitude.com/mcp"
    }
  }
}

About this resource

Content

Amplitude MCP Server is Amplitude's official remote Model Context Protocol server for bringing product analytics into AI clients. It lets Claude Code and other MCP-capable tools query Amplitude content, inspect product behavior, work with experiments, and create or edit Amplitude resources through natural language.

The server is useful for teams that already rely on Amplitude dashboards, experiments, cohorts, session replay, and feedback workflows. Instead of asking Claude to reason from pasted screenshots or stale exports, users can connect the assistant to live Amplitude context through OAuth and the existing Amplitude permission model.

Amplitude documents two product-data endpoints:

  • US/default: https://mcp.amplitude.com/mcp
  • EU residency: https://mcp.eu.amplitude.com/mcp

Use the EU endpoint only when the relevant Amplitude data is in the EU region.

Features

  • Remote Streamable HTTP MCP server with OAuth authentication.
  • Product analytics questions over Amplitude projects the user can already access.
  • Chart, dashboard, and notebook search, retrieval, analysis, creation, and editing workflows.
  • Experiment and feature-flag workflows for setup, monitoring, stale-test checks, and result interpretation.
  • Session Replay and debugging workflows for finding replay examples and summarizing user friction.
  • Feedback and opportunity analysis across product signals.
  • AI agent analytics workflows for monitoring topics, quality, cost, errors, and low-quality sessions.
  • Cohort, metric, account health, and journey-comparison workflows.
  • Progressive tool discovery with ?discovery=progressive to keep the initial tool list small and discover schemas on demand.
  • Organization admin controls for allowing or blocking MCP access.

Use Cases

  • Ask Claude why a weekly active user metric dropped and have it search related charts, experiments, dashboards, and session replays.
  • Build or update an executive product dashboard from a concise set of goals.
  • Investigate signup, activation, checkout, onboarding, or retention funnels without manually copying Amplitude screenshots into a prompt.
  • Monitor running experiments and flag stale tests, missing metrics, or risky rollout decisions.
  • Create a cohort of high-value users, churn-risk users, or users who completed a specific behavior in the last 30 days.
  • Summarize dashboard or notebook context before a product, growth, support, or executive meeting.
  • Review AI agent quality, cost, performance, topics, and error trends when those signals are tracked in Amplitude.

Installation

Claude Code

Add the standard Amplitude MCP endpoint as a remote HTTP server:

claude mcp add -t http -s user amplitude https://mcp.amplitude.com/mcp

Open the MCP connection in Claude Code and complete the Amplitude OAuth flow. Make sure you authenticate with the right Amplitude organization and project.

For EU-resident Amplitude data, use the EU endpoint instead:

claude mcp add -t http -s user amplitude-eu https://mcp.eu.amplitude.com/mcp

Progressive discovery

For large organizations or token-sensitive workflows, use progressive discovery:

claude mcp add -t http -s user amplitude https://mcp.amplitude.com/mcp?discovery=progressive

Then ask Claude to call get_context, list_tool_categories, get_category_tools, and describe_tool before using a specific analytics, dashboard, experiment, cohort, or replay tool.

Security Notes

Amplitude MCP uses the current user's Amplitude permissions, but the data it returns still enters the connected AI client and model session. Treat it as a production data integration. Keep access limited to approved users, use admin controls when AI access is not allowed, and require confirmation before any write-capable tool creates or changes Amplitude content.

Related Resources

#amplitude#mcp#analytics#experiments#dashboards#oauth

Source citations

Signals

Loading live community signals…

More like this, weekly

A short, calm digest of reviewed Claude resources. Unsubscribe any time.