Peek
Feed locally recorded rrweb browser sessions—DOM, console, and network—into your agent so it can reproduce and fix UI bugs without live browser guesswork.
Overview
peek-mcp is a Ship-phase MCP server that exposes locally recorded rrweb browser sessions—DOM, console, and network—to AI coding agents.
What is this MCP server?
- Exposes rrweb DOM replay context to AI coding agents via MCP
- Includes console logs and network activity from recorded sessions
- Runs fully local—session data stays on your machine
- npm package @peekdev/mcp with npx runtime hint (0.1.0-alpha.15)
- Part of Cubenest rrweb-stack; website at peek.cubenest.in
- Package version 0.1.0-alpha.15
- Three signal classes called out: rrweb DOM, console, network
- Fully local processing (no cloud session requirement in server description)
Community signal: 3 GitHub stars.
What problem does it solve?
You have a session recording but still spend hours re-explaining clicks, errors, and network failures to your agent without structured replay data.
Who is it for?
Frontend-heavy solo builders who record rrweb sessions and want Claude Code or Cursor to analyze repros offline.
Skip if: Teams that need live remote browser control, cloud session hosting, or backend-only services with no web UI.
What do I get? / Deliverables
After registering peek-mcp, your agent can query recorded session artifacts locally and debug UI issues with DOM, console, and network context in one place.
- MCP tools that surface DOM, console, and network data from recorded sessions
- Agent-driven debugging workflows that do not upload session files to the cloud
Recommended MCP Servers
Journey fit
Recorded-session review is a Ship concern: you validate behavior, trace failures, and tighten fixes before users hit the same breakage in production. Testing fits because peek turns passive session captures into evidence your agent uses while debugging regressions and flaky flows.
How it compares
Local session-replay MCP bridge, not a hosted observability suite or a Playwright test runner skill.
Common Questions / FAQ
Who is Peek-mcp for?
Indie web developers and small teams who capture browser sessions with rrweb and want their AI agent to inspect those recordings during bug fixing.
When should I use Peek-mcp?
Use it in Ship when you are testing or triaging UI bugs and have a recorded session with DOM, console, or network signals you want the agent to read directly.
How do I add Peek-mcp to my agent?
Add the @Peekdev/mcp stdio server via npx in your MCP settings for Claude Code or Cursor, following the rrweb-stack Peek-mcp package docs on GitHub.