← Back to Discover

Poirot

Native macOS app for browsing Claude Code sessions, with transcripts, diffs, usage analytics, usage-limit gauges and config views. Reads local transcripts, and the README says it works offline.

Coding agentsObservability & evals

What it does

Poirot is a SwiftUI desktop app that reads the transcripts Claude Code stores on your machine and shows them as a browsable history. Sessions are grouped by project, with a full conversation timeline, rendered markdown, syntax-highlighted diffs, collapsible tool blocks and thinking blocks, and you can click a shell command to copy it or open it in your terminal. Beyond history, it includes a dashboard of token use, cost breakdowns and model distribution, plus gauges for subscription rate limits in the dashboard and menu bar. Other screens cover slash commands, skills, MCP servers, sub-agents, plugins, hooks, plans, memory files, prompt history and debug logs, with fuzzy search across all of them. The README says it needs no login, has no tracking or analytics, adds no extra cost, and works offline. It requires macOS 15 or later, can be installed with Homebrew, and sessions can be exported as Markdown or PDF. The project is MIT licensed.

Capabilities

  • Session browser grouped by project with a full conversation timeline
  • Syntax-highlighted diffs and collapsible tool and thinking blocks
  • Token, cost and model analytics plus rate-limit gauges
  • Create and edit sub-agents, manage skills and hooks, and browse MCP servers and plans
  • Fuzzy search across sessions, commands, memory and logs
  • Export sessions as Markdown or PDF

When to use it

  • You use Claude Code and want a readable history of past sessions and their changes
  • You want to track token use, cost and rate-limit headroom in one place
  • You want to manage Claude Code sub-agents, skills and hooks from a visual interface
  • You prefer a local app with no login or telemetry