Skip to content

Amplitude Agent Skills - Setup Guide

Source: amplitude/mcp-marketplace (42 stars, MIT LICENSE; actively maintained - pushed Oct 8, 2026) Skill family: amplitude/mcp-marketplace (39 SKILL.md files in-repo; 38 indexed listings) Combined Installs: ~175,215 across indexed listings (Oct 8, 2026 snapshot) Category: Analytics / Product Data Quality Tier: Production (official Amplitude publisher; MIT LICENSE in-repo; active development - verified Oct 8, 2026)

The 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." It works with Claude Code, Cursor, Codex, Gemini CLI, and Claude. The roster covers the full product-data loop: discover where events can be captured, plan and diff instrumentation changes, implement events against a tracking plan (taxonomy, naming conventions, data-quality audits, funnel design, event deprecation policy), then analyze charts, dashboards, experiments, cohorts, and user journeys, with daily and weekly briefs for ongoing reporting. A first-run onboarding skill handles US vs EU data residency, Docs MCP setup, and project access, and event-description-generator shows an orchestrator pattern that delegates chunk work to worker subagents. 38 skills are indexed on skills.sh; six flagship workflows carry nearly all of the installs so far.

Installation

# skills.sh - install the full suite
npx skills add amplitude/mcp-marketplace

# Or install a single skill
npx skills add amplitude/mcp-marketplace --skill taxonomy

Claude Code users can install the plugin directly:

claude plugin install amplitude

Or from inside a Claude Code session:

/plugin install amplitude
/reload-plugins

Roster - Indexed Skills

Six flagship skills carry the suite (installs from the Oct 8, 2026 skills.sh snapshot):

Skill Domain Installs What It Does
discover-event-surfaces Event discovery 29,271 Scans a codebase for surfaces where events can be captured before instrumentation is added
diff-intake Change planning 29,262 Intakes and diffs instrumentation changes so tracking edits stay reviewable
discover-analytics-patterns Analytics discovery 29,259 Finds analytics patterns in an app to guide instrumentation and reporting
instrument-events Instrumentation 29,202 Implements event instrumentation in the codebase
taxonomy Taxonomy / governance 29,101 "Taxonomy Generation & Data Auditing": tracking plans, event and property naming conventions, data-quality audits (duplicates, stale events, missing metadata), funnel design, event deprecation policy, AI-readiness
add-analytics-instrumentation Instrumentation 28,969 Adds analytics instrumentation to a product surface

Long tail (32 skills, 1-6 installs each): the analysis, monitoring, and reporting class - create-chart, create-dashboard, analyze-chart, analyze-dashboard, analyze-experiments, monitor-experiments, compare-user-journeys, user-cohort-forensics, live-data-forensics, replay-ux-audit, debug-replay, diagnose-errors, daily-brief, weekly-brief, scheduled-report-refresh, tracking-plan-audit, and more. Two standouts: event-description-generator (5 installs) orchestrates and delegates chunk work to worker subagents, and getting-started-with-amplitude (in the repo tree; not in the indexed roster) covers first-run onboarding: US vs EU data residency, Docs MCP, and project access.

Why This Matters for Hermes Agents

Product-analytics instrumentation and reporting is a recurring operator workflow: wiring events correctly, keeping tracking plans clean, and turning charts and experiments into decisions. This suite keeps that loop inside the agent session - discover event surfaces, plan and diff instrumentation changes against a tracking plan, then pull daily or weekly briefs without a human re-explaining metric definitions each time. Clean tracking plans also matter because agents read the same event names humans do: a consistent taxonomy is what keeps product data queryable by the automations built on top of it.

Usage and Triggers

Action Prompt / Command
Plan instrumentation "Find where signup and checkout events should be instrumented in this codebase"
Audit taxonomy "Review our tracking plan for naming-convention violations, duplicate events, and missing metadata"
Instrument events "Implement the events in this tracking plan and show me the diff before applying it"
Analyze experiments "Analyze the results of our latest A/B experiment and summarize the winner"
Report "Generate this week's product brief from our Amplitude dashboards"

Verification

# Spot-check the source repo (expect: 42 stars, pushed 2026-10-08, MIT)
curl -s https://api.github.com/repos/amplitude/mcp-marketplace | grep -E '"stargazers_count"|"pushed_at"'

After installing, start with an audit-style workflow - run the taxonomy skill against a sample tracking plan and review its output - before letting the agent touch production instrumentation. No live install test was performed for this guide; counts and verdicts are the Oct 8, 2026 skills.sh snapshot.

Security

skills.sh per-skill verdicts (sampled: taxonomy, verified Oct 8, 2026) - verdicts vary per skill; check the skill's security page on skills.sh before production use:

Skill Gen Agent Trust Hub Socket Snyk
taxonomy Pass Pass Warn

Integration with CorpusIQ

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations

  • The repo is named mcp-marketplace while the Amplitude plugin content ships under plugins/amplitude/skills/ (37 skills), plus one skill each under plugins/amplitude-grok/ and plugins/amplitude-experimental/. The tree carries 39 SKILL.md files against 38 indexed listings, so small index/repo drift is normal.
  • Install distribution is top-heavy: the six flagship skills carry 175,064 of the 175,215 combined installs; the remaining 32 sit at 1-6 installs each, so treat the long tail as new rather than field-proven.
  • First seen in the Mar 31, 2026 sweeps; all numbers here are the Oct 8, 2026 snapshot and will drift.
  • Only a subset of skill bodies was reviewed for this guide (taxonomy, event-description-generator, getting-started-with-amplitude); other descriptions derive from skill names and the README.
  • Instrumentation skills change real product code: run them on a branch and review diffs before merging.

Pro Tips

  1. Start with the flagships: discover-event-surfaces then diff-intake then taxonomy is the natural order for a fresh instrumentation project; review each output before moving on.
  2. Run taxonomy before generating new events - it audits duplicates, stale events, and missing metadata first, so you instrument only what the plan actually needs.
  3. event-description-generator delegates chunk work to worker subagents; scope each chunk to one module or file and verify the merged output.
  4. The long tail is where monitoring and reporting live (daily-brief, weekly-brief, monitor-experiments); install the full suite rather than just the flagships.
  5. Install counts lag reality: when a long-tail skill matches your workflow, read its SKILL.md and test it instead of dismissing it on its 1-6 install count.

← Skills Catalog | Skills Marketplace → Powered by CorpusIQ

AI answers on your data. See where every number comes from.

Connect your tools and ask in ChatGPT, Claude, or Perplexity. Enterprise-grade answers in minutes. From $29.95/month.

Start free, no credit card
Contact Terms Privacy Pricing