PostHog AI Observability

What it does

PostHog AI Observability records LLM traces, including generations, spans, tool calls, tokens, latency, and cost, and attaches them to users. Traces can be opened in a timeline view and linked to PostHog session recordings. Evaluations score live traffic, and anomaly alerts flag unusual cost, latency, or error patterns.

People use it to debug conversations span by span, compare costs and latency across models, and check spend before and after deploys. Engineers can also query traces from an editor such as Cursor or VS Code through the PostHog MCP server. Its distinct feature is connecting AI traces with product analytics, error tracking, session replay, and experiments in the same product.

It suits AI engineers, product engineers, PMs, and finance teams already using PostHog or evaluating an all-in-one analytics stack. Getting started requires a free PostHog account; no credit card is needed for the free allowance. Some capabilities are marked Beta or Partial, and the page does not list every supported provider or confirm rates beyond the free allowance.

What it can do

• Trace LLM conversations end to end with span timelines, token counts, and automatic cost calculation
• Run LLM-as-a-judge and code evaluators on live traffic and datasets
• Detect anomalies in LLM cost, latency, and errors with learned baselines
• Link traces to session recordings and user profiles
• Query LLM costs and traces from editors through the PostHog MCP server
• Manage, version, and A/B test prompts with deployment via API

Works with

• PostHog MCP server (Claude, Cursor, VS Code, Windsurf, Codex, Zed, Lovable, Replit, v0)
• OpenTelemetry trace ingestion

Before you choose

• Some features (such as LLM translation, sentiment classification, human annotation, and agent reports) are marked Beta or Partial.
• Privacy mode can exclude sensitive data, and prompts and responses can be masked before storage; configure these if handling sensitive content.
• Free allowance of 100K AI Observability events per month is shared with evaluations; usage beyond it is billed per event.

Listed under

• Coding
• DevOps and observability

Get started

Sign up free; the free allowance does not require a credit card. Installing with AI uses a wizard command.

1. Create a free PostHog account at app.posthog.com/signup.
2. Install the SDK for your stack, or run the install wizard: npx -y @posthog/wizard@latest ai-observability.
3. Optionally add the PostHog MCP server to your editor with npx @posthog/wizard mcp add to query traces from your coding agent.

Free allowance of 100K AI Observability events per month, shared with evaluations, which are billed with AI observability. Beyond that, the AI Observability page states pricing starts at $0.00035/event and reduces with volume to $0.00006/event. Check the product's rate table for current rates.

Source: posthog.com/ai-observability, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• AI Observability – LLM tracing, evals, and costs | PostHog ↗posthog.com
• PostHog pricing – Start free, scale when you need ↗posthog.com

Company website

posthog.com: product, pricing and policy pages

Hacker News

Launch posts and discussion

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.