MCP Analytics

Contents

MCP Analytics shows how AI agents use your MCP tools. It records tool calls, agent intent, reported models, failures, and requests for missing capabilities. It supports PostHog's hosted MCP server and servers that use the MCP Analytics SDKs.

Each tool invocation creates a $mcp_tool_call event in your standard events table. Use these events in MCP Analytics, Product Analytics, or SQL. Self-driving uses them to find problems and create reports. It can open a pull request for a server you own.

MCP Analytics is in beta

The in-app MCP Analytics views are in beta. The @posthog/mcp SDK is published as a 0.x release, so event names, properties, and tracing behavior may change before 1.0. Pin a version and don't depend on it for production reporting yet.

Get started

Where you can use it

You explore MCP usage in the PostHog web app. The other surfaces let you query the same data, act on it, and pull it into your own tools.

PostHog WebBeta

Dashboards, session replay for agents, per-tool quality, and intent clustering.

Explore MCP usage

PostHog MCP

Query sessions, tool stats, failures, and intent clusters from any MCP client.

Query over PostHog MCP

PostHog DesktopBeta

Read Self-driving reports about failing tools, and merge the fixes for servers you own.

Fix tools in PostHog Desktop

API

Pull sessions, tool calls, intent clusters, and feedback into your own systems.

Use the API

Where its data comes from

MCP Analytics runs on the events your MCP server sends to PostHog. The SDK emits them automatically, and PostHog's hosted MCP server is instrumented the same way.

Tool calls

The canonical $mcp_tool_call event: tool name, client, latency, errors, and session.

Event reference

Lifecycle and discovery

Legacy initialize handshakes, tools/list, and per-request protocol versions power client and spec adoption breakdowns.

See all events

Agent intent

What the agent said it was trying to do, clustered into themes across sessions.

Capture intent

Missing capabilities

Gaps agents report through the get_more_tools virtual tool when the right tool doesn't exist.

Track missing capabilities

How MCP Analytics works with Self-driving

Self-driving uses your tool calls to find tools that need attention. A scout checks for high failure rates, repeated calls, slow responses, and large responses. It considers call volume and reach, then creates reports grouped by the team that owns the tools.

For your own server, the report includes a fix loop metric: the value that the proposed fix should improve. A coding agent then opens a pull request. For PostHog's hosted server, the report goes to a human without an attached repository. This path does not open a pull request.

Tool callsEvery agent invocation of your MCP tools
SignalsFailures, retries, and slow responses surfaced by a scout
One reportGrouped by owning team, with the evidence behind it
Draft PROnly when you own the server and the fix is concrete
You mergeNothing ships without a human

Select Create fix task in the tool quality view to start a fix in a repository you choose. See PostHog Desktop for the report and fix process.

Still have questions?

Was this page useful?