Amplitude AI Plugin

Official Amplitude plugin for AI coding tools. Turn your AI assistant into a product analyst — instrument analytics, analyze charts, run experiments, and understand users directly from your editor.

Works with Claude Code, Cursor, Codex, Gemini CLI, and Claude.

Installation

Claude Code

claude plugin install amplitude

Or from inside Claude Code:

/plugin install amplitude
/reload-plugins

Then authenticate when prompted.

Cursor

Install from the Cursor Marketplace or add manually in Cursor Settings > Plugins.

Codex

Add the marketplace:

codex plugin marketplace add amplitude/mcp-marketplace

Install the plugin from inside Codex:

codex
# Then run /plugins, select Amplitude, and install
/plugins

Gemini CLI

gemini extensions install https://github.com/amplitude/mcp-marketplace

Local development

To work on the plugin from a local checkout in Claude Code:

git clone https://github.com/amplitude/mcp-marketplace
claude --plugin-dir ./mcp-marketplace/plugins/amplitude

Then run /mcp from inside Claude Code, select plugin:amplitude:amplitude, and follow the browser prompts to authenticate.

What's Inside

Plugin	Description
amplitude	Reusable analysis and instrumentation skills covering charts, dashboards, experiments, session replays, reliability, AI agent analytics, and analytics tracking workflows

Amplitude Plugin

The amplitude plugin turns your AI assistant into an expert product analyst and instrumentation partner. Skills are organized into seven areas:

Core Analytics

Skill	What it does
`create-chart`	Creates Amplitude charts from natural language descriptions
`create-dashboard`	Builds dashboards from requirements, organizing charts into logical sections
`analyze-chart`	Deep-dives a chart to explain trends, anomalies, and likely drivers
`analyze-dashboard`	Reviews a dashboard end-to-end, surfacing key takeaways and areas of concern

Product Insights

Skill	What it does
`analyze-experiment`	Designs A/B tests, monitors running experiments, and interprets results
`monitor-experiments`	Triages all active and recently completed experiments by importance
`analyze-feedback`	Synthesizes customer feedback into themes — feature requests, bugs, pain points, praise
`analyze-account-health`	Summarizes B2B account health with usage patterns, risk signals, and expansion opportunities
`discover-opportunities`	Finds product opportunities by cross-referencing analytics, experiments, replays, and feedback
`compare-user-journeys`	Compares two user groups side-by-side to surface behavioral differences

Session Replay & Debugging

Skill	What it does
`debug-replay`	Turns bug reports into numbered reproduction steps by extracting the interaction timeline from Session Replay
`replay-ux-audit`	Watches multiple session replays for a flow and synthesizes a ranked friction map
`diagnose-errors`	Triages product issues across network failures, JS errors, and error clicks
`monitor-reliability`	Proactive reliability report from auto-captured error data so issues surface before users complain

AI Agent Analytics

Skill	What it does
`analyze-ai-topics`	Analyzes what users ask AI agents about and how well each topic is served
`investigate-ai-session`	Deep-dives specific AI agent sessions or failure patterns for root-cause analysis
`monitor-ai-quality`	Delivers a proactive health report on AI agents covering quality, cost, performance, and errors
`review-agent-insights`	Synthesizes recent results from Amplitude's AI agents into a unified, impact-ranked narrative

Analytics Instrumentation

Skill	What it does
`diff-intake`	Reads a PR or branch diff and outputs a structured `change_brief` YAML for downstream skills
`discover-event-surfaces`	From a change brief, lists candidate analytics events for PM prioritization
`discover-analytics-patterns`	Maps how analytics is already implemented in the repo (SDK calls, naming, imports)
`instrument-events`	From prioritized event candidates, builds a concrete instrumentation plan and JSON tracking plan
`add-analytics-instrumentation`	End-to-end workflow — reads code, decides what to track, and produces a full instrumentation plan in one pass
`taxonomy`	Source of truth for event taxonomy generation, data auditing, and governance best practices

A typical flow: diff-intake → discover-event-surfaces → instrument-events, with discover-analytics-patterns ensuring new tracking matches existing conventions.

Briefings

Skill	What it does
`daily-brief`	Morning briefing of the most important changes across your Amplitude instance
`weekly-brief`	Weekly recap of trends, wins, and risks to share with your team or leadership

Bonus

Skill	What it does
`what-would-lenny-do`	Answers product strategy questions by searching Lenny Rachitsky's archive (requires `lennysdata` MCP)

Repository Structure

.agents/plugins/
  marketplace.json            # Marketplace catalog (Codex)
.claude-plugin/
  marketplace.json            # Marketplace catalog (Claude Code)
.cursor-plugin/
  marketplace.json            # Marketplace catalog (Cursor) — kept in sync with .claude-plugin
plugins/
  amplitude/
    .codex-plugin/
      plugin.json             # Plugin manifest (Codex)
    .claude-plugin/
      plugin.json             # Plugin manifest (Claude Code)
    .cursor-plugin/
      plugin.json             # Plugin manifest (Cursor) — kept in sync with .claude-plugin
    .mcp.json                 # MCP server config shared by Codex and other clients
    skills/
      add-analytics-instrumentation/
      analyze-account-health/
      analyze-ai-topics/
      analyze-chart/
      analyze-dashboard/
      analyze-experiment/
      analyze-feedback/
      compare-user-journeys/
      create-chart/
      create-dashboard/
      daily-brief/
      debug-replay/
      diagnose-errors/
      diff-intake/
      discover-analytics-patterns/
      discover-event-surfaces/
      discover-opportunities/
      instrument-events/
      investigate-ai-session/
      monitor-ai-quality/
      monitor-experiments/
      monitor-reliability/
      replay-ux-audit/
      review-agent-insights/
      taxonomy/
      weekly-brief/
      what-would-lenny-do/

Requirements

MCP-compatible client – Claude Code, Cursor, Codex, or Claude
Amplitude account with API access

Contributing

We welcome contributions! Whether it's a new skill or improvement:

Fork this repo
Create a new branch for your feature
Follow the existing patterns in plugins/ for structure
Submit a PR with a clear description

Resources

License

MIT

Support

Issues: GitHub Issues
Amplitude: amplitude.com

Amplitude

Installation

How to install

README

Amplitude AI Plugin

Installation

Claude Code

Cursor

Codex

Gemini CLI

Local development

What's Inside

Amplitude Plugin

Core Analytics

Product Insights

Session Replay & Debugging

AI Agent Analytics

Analytics Instrumentation

Briefings

Bonus

Repository Structure

Requirements

Contributing

Resources

License

Support

You might also like