MCP Analytics dashboard showing tool calls, error rates, and latency
Two hedgehogs holding puzzle pieces that fit together
Beta

See everything agents do on your MCP server

“Summarize the agent intents behind my top 5 tools.”

Get started - free

Install with AI in a single prompt

Paste into your terminal or code editor and make AI do the work.

Learn more

MCP•Watch a demo•Talk to a human

What does it do?

MCP Analytics captures every tool call agents make to your MCP server: what they sent, what came back, how long it took and whether it failed. We ask agents why they're calling each tool, so you see the user's goal alongside the call. Sessions tie the calls together, so one agent task reads as a single story.

Who is it for?

Role
Use cases
Engineers building MCP servers
Find and fix broken or slow tools. Get error rates and p95 latency by tool. Monitor health and performance.
Engineers building MCP tools
Find the slow and failing tools and send a coding agent to fix them.
Engineers maintaining APIs
Identify and fix endpoints causing high error rates in tool calls.
Product managers
Learn who uses the server and what those users try to do.

How do I use it?

Ask your agent about your MCP server

Don't want to open a dashboard? Query MCP analytics from Claude Code, Cursor, or any client. Ask for this week's errors and slowest tools, then fix them in the same session.

Query MCP tool stats
Ask which tools are used, and if they're healthy. Check a tool's volume and confirm that events arrive after a deploy.
Know client and harness breakdown
Know which clients are calling and which models are used the most to adapt your MCP server to what your customers are using.
Dig into tool quality
Check reliability, p95 latency and sort a failing tool's errors into groups to find the cause before a fix.

Key features

MCP Analytics dashboard showing tool calls, error rates, and latency

Every tool call

Each tool call becomes an event with response, duration, and errors. See exactly how many tool calls you run and which ones fail.

The call
Tool name and description, sanitized arguments and response, and duration in milliseconds
Clients
Know who's calling with details on harness, user agent, and client name
Models
Know which models are most used from client metadata or the agent's own report
Failures
Error type and message on every failed call, plus an $exception event so the tool shows up in Error tracking

AI prompts

Ask PostHog AI, or your agent through the PostHog MCP server. Start with the usual questions about usage and errors, then go deeper: exact error messages, release checks, clients, and the tools agents never call. Already signed in? Click a prompt to try it.

Works with other PostHog tools

Use MCP Analytics with these other PostHog apps to maximize shareholder value.

And the rest of PostHog:

Install

Add the MCP Analytics SDK to your server and wrap it once. The wizard sets up TypeScript and Python servers for you. If you'd rather one-shot it with your own coding agent, every install page has a Markdown version for agents.

Learn more

Get started – free

No credit card required.

Install with AI

Run this command in your terminal or AI editor.
npx -y @posthog/wizard@latest mcp-analytics

Supports Next.js, React, Python, and

Install without AI

Sign up for a free account and follow the install instructions for your stack.

Get started - free