Replay MCP
Quickstart
Learn how to use Replay MCP with your AI agent.
This guide walks through the fastest path to debugging with Replay MCP: install skills, connect the MCP server, record your app, and let your agent investigate.
Other ways to record
This quickstart uses an agent-driven CLI recording. You can also record with the Chrome Extension or Playwright in CI. See How to Record for a comparison of all methods.
Install the Replay Skills
Install the Replay skills so your agent knows how to use the recording and debugging tools:
Terminalnpx skills add https://github.com/replayio/skills --skill '*'
Install the Replay MCP Server
Terminalclaude mcp add replay\--transport http https://dispatch.replay.io/nut/mcp\--header "Authorization: YOUR_API_KEY"
Refer to these docs to see how you can generate an API key.
Record your application
Instruct your agent to open replay-cli. This will open the browser that records the activity. Replicate the issue that you want to debug. Once replicated, exit the browser completely.
Example prompt:
Please help me debug issues on my application.
Use replay-cli skill to record my application running on localhost:3000.
Then use Replay MCP to debug the recording.
Usually this is enough for Replay to debug your issue and provide your agent with enough information. The instruction will:
- Record the activity
- Notice when the browser is closed
- Upload the recording to Replay DevTools (each recording has a unique id )
- Call the Replay MCP server to fetch data from the recording
- Your agent will then take care of the rest thanks to the analysis provided by Replay MCP
For best results, keep your recordings short.
Use Replay MCP to debug your application
With the instructions above, Replay MCP should trigger automatically. It will fetch analysis data and provide it back to your agent. The agent usually goes ahead and proposes a fix for the issue.
What's next?
- Tools Reference — see all the tools your agent can use when investigating a recording
- MCP Overview — learn about connection modes, skills, and tips for effective debugging