Find the span that broke, then fix it

Install with AI in a single prompt

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

What does it do?

A trace times one request through every service it touches. Open the waterfall and there’s the slow query six services deep, or the async job a stack trace gives up on. Traces come in over the same OpenTelemetry setup as Logs, so one install covers both. Traces get you to where a request went wrong, and the logs from that span tell you why.

Who is it for?

Role

Use cases

Backend Engineers

Follow one request across every service it touches to find the span that actually spent the time, including async work a stack trace never shows

Product Engineers

Explain the slow page analytics flagged, down to the query behind it, without guessing which service to open first

Platform / DevOps

Watch latency and request volume per service, and catch the p99 regressions you can't reproduce locally

Support Engineers

Take a trace ID from a complaint and see exactly where that request stalled, then jump to the session replay beside it

AI-assisted teams

Let scouts and coding agents read traces over MCP to locate a regression and open the pull request

How do I use it?

Read traces from your editor

The PostHog MCP server gives coding agents direct access to your spans. Ask your agent which service is slow, have it pull the trace behind a request, and fix the span it points at, without leaving your editor.

Query and count spans

Filter by service, duration, status, and attributes, then pull counts or a sparkline to see how a regression is trending.

Pull a whole trace

Retrieve every span under a trace ID so the agent reads the same waterfall you would.

Find where the time goes

Aggregate the call tree, break latency down by attribute, and check a duration histogram or latency heatmap.

Discover what's available

List the services emitting spans and the attributes they carry, with the values you can filter on.

Install the PostHog MCP

Learn more

npx @posthog/wizard mcp add

Supports Next.js, React, Python, and

Top features

Every span of a request, in one waterfall

One request becomes a tree of spans across every service, queue, and third-party call it touches.

Follow the trace_id to see where the time went and which span actually broke, even across async boundaries, where stack traces give up.

A distributed trace shown as a waterfall of spans

AI prompts

Query spansquery-apm-spans

Services and attributesapm-services-list

Install

Tracing is built on OpenTelemetry, so any OTLP-compatible exporter works. There are no PostHog-specific tracing packages to adopt. Point your existing OpenTelemetry setup at PostHog and spans start arriving.