Replay Vision - PostHog

Replay Vision

Your product, watching itself

Replay Vision reads through a filtered set of recordings, tells you in plain language what went wrong, and opens a pull request with the fix. The problem fixes itself for you. You just hit merge.

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

The old way to use session replay

Session replay tells you what happened, but only if you press ▶ Play on session after session after session.

Using replay(cir. 2022-2026)

1. Notice something's offHuman

2. Watch hours of sessionsHuman

3. Spot the patternHuman

4. Diagnose the causeHuman

5. Write the fixHuman

6. ShipHuman

You're the one who has to watch them all, spot the pattern, and figure out the fix – the tool just hands you the footage.

The PostHog way

Replay Vision watches your session recordings for you. Describe what to look for once, and a scanner reads every matching session, video and events, and turns what it sees into structured observations you can query, chart, and alert on.

The loop(cir. 2026–)

Replay Vision watches the sessionsMachine

Observation what happened, as dataMachine

Signal raised to your InboxMachine

Agent opens the PRMachine

You review & mergeHuman

The friction you'd never have watched becomes a signal in your Inbox – where an agent picks it up and opens the pull request. You just hit Merge ↵.

Who is it for?

Replay Vision is used across teams depending on your role.

Role

Use cases

Product Engineers

Scan sessions in bulk for the failure you can't reproduce locally

PMs & Designers

Score friction and spot dead ends across a release without watching a replay

Growth Engineers

Find where users bleed out of onboarding and signup funnels

Support & UX Research

Classify what users were actually trying to do, at scale

Founders

Skim a one-line summary of every session instead of spending hours watching them

How do I use it?

There are a few ways to put Replay Vision to work.

Author, run, and read scanners without leaving your editor

Your AI coding agent can call Replay Vision directly through the PostHog MCP – in Cursor, Claude Code, Codex, VS Code, or any MCP client.

Author a scanner

Describe what to look for in plain language; the agent drafts the config, projects the monthly volume, and creates it.

Investigate a session

Scan a specific recording on demand and pull the result – plus the model's reasoning – straight into context.

Research before coding

Read a scanner's recent observations to see how users actually behave before you change anything.

Audit a scanner

List recent observations to decide whether the prompt or filters need tuning.

Install the PostHog MCP

Learn more

npx @posthog/wizard mcp add

Supports Next.js, React, Python, and

Top features

What it looks for

Replay Vision runs scanners – AI probes you configure and point at your sessions. Pick a type, describe what to look for, and it produces structured output on each session it scans. Start from a built-in template or from scratch.

Create from scratch

Build a fully custom scanner – pick a type and write your own prompt and config.

Dead ends (Monitor)

Catch the moment someone hits a wall: scrolling, hovering with no CTA, then rage-quitting.

Session summary (Summarizer)

The TL;DR of the session, so you don't sit through 14 minutes of someone scrolling.

User intent (Classifier)

Classify the session by what the user appeared to be trying to do.

Frustration score (Scorer)

Rate how much friction a page caused, on a scale you define.

Session outcome (Classifier)

Tag what actually happened – task completed, abandoned, errored, and so on.

Every observation comes with a confidence score and citations that link straight to the exact moment in the recording, so you can verify it in one click.

AI prompts

Ask your AI agent to author scanners, scan a session, or read what Replay Vision found – without leaving your editor. Works in any MCP client via the PostHog MCP: Cursor, Claude Code, Codex, VS Code, and more. Already signed in? Click a prompt to try it.

Create a scannervision-scanners-create

Size it before you commitvision-scanners-estimate-create · vision-quota-retrieve

Scan a session on demandvision-scanners-scan-session

“Scan session abc123 with the 'Dead-end pages' scanner and tell me what it found, including the reasoning.”

Read observationsvision-observations-list · vision-scanners-observations-list

Tune a scannervision-scanners-list · vision-scanners-update

Works with other PostHog tools

Use Replay Vision with these other PostHog apps to get more out of every session.

Session Replay
The recordings your scanners watch; point Replay Vision at any set of sessions you're already capturing.
Product Analytics
Every observation lands as a $recording_observed event, so you can trend and break down what scanners find like any other insight.
Dashboards & Alerts
Put frustration scores or dead-end counts on a dashboard, and get alerted when a scanner's output crosses a threshold.

Questions?

Answers

There are a few ways you can get answers to specific questions about Replay Vision.

Check the docs
We have an entire docs-wizard dedicated to docs gardening.
Search the community forums
With over 400 discussions about Replay Vision, there's a good chance your answer is already answered!
Ask PostHog AI
It's an incredibly useful AI chat that understands the product, docs, community questions, our codebase, GitHub issues, and industry knowledge.
Talk to a human
Dedicated humans are standing by and ready to assist. Best for questions about volume pricing, terms, and sexy legal stuff.