
Jittle Lamp
QA evidence that AI agents can read, write and run.
Records browser sessions as reviewable evidence: video, user actions, console and full network traffic on one timeline. Test cases are plain-language transcripts; the first run is agentic, later runs replay from cache without the model.
- Browser extension and desktop recorder, secrets masked before upload
- Agent-run tests that replay as Playwright code
- MCP server with 40+ tools for Claude Code and Codex
- TypeScript
- Rust
- Playwright
- MCP





