Poirot Updates: Multi-Agent Visibility, Sessions Navigation, MCP Wizard, Menu Bar Companion
Poirot shipped four features this week. Every single one was built by autonomous coding agents running on the project board.
See all sub-agents spawned during a Claude Code session. Track which agents worked on what, their tool calls, and how they coordinated.
If you’re running complex multi-agent workflows, this is how you debug them. You get full visibility into the agent tree, what each agent did, and how long it took.
Browse all your Claude Code sessions in a proper navigation view. Sessions are grouped by project with collapsible sections, search, and model/token badges.
The 3-column layout shows the session list, detail view, and inspector side by side. No more hunting through files to find the session you want.
Add MCP servers to your Claude Code config without editing JSON by hand.
A catalog of 15 popular servers is included out of the box: GitHub, Playwright, Slack, Notion, PostgreSQL, and more. Each one comes with guided setup for environment variables and arguments. You can also add custom servers and edit existing ones.
No more typos in claude_desktop_config.json.
Poirot now lives in your menu bar. Quick access to sessions, stats, and search without opening the full app window. One click to jump into any session.
What I find most interesting about this batch is how it was built. These features were planned on a GitHub project board, picked up by autonomous coding agents, implemented, tested, and PR’d without manual intervention.
The cycle from issue to merged PR runs on its own. I review the PRs, but the implementation is fully autonomous.
Open source, native macOS, under 6 MB.
brew install --cask poirot