Tomarigi
macOS app showing each Claude Code or Codex session as a bird; Jev optionally flags abusive messages and plain-text questions.
Overview
Tomarigi is a macOS desktop app that watches AI coding agent sessions from Claude Code and Codex and shows each one as a bird (or other icon) in an always-on-top window. At a glance you can see which session is working, waiting for a reply or done, and clicking a bird jumps to the Ghostty pane running it. It reads transcripts from ~/.claude/projects and ~/.codex/sessions, so no hooks or config changes are needed.
Jev is an optional, bring-your-own-key feature. With a TypeSafe key, each message you send is checked for abuse toward the AI, and the bird gets an anger mark if it looks abusive; Jev also spots questions the agent asked in plain text. The README notes the mark judges one message at a time and is not a policy verdict. OpenAI or Anthropic keys can optionally summarize finished turns.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this message abusive toward the AI?(inferred) | yes · no |
| Yes / no | Is the agent asking the user a question in plain text?(inferred) | yes · no |
How it works
- Read agent transcripts from local session folders.
- Show each session's status as a bird in a floating window.
- With a TypeSafe key, send each user message to Jev for an abuse check.
- Display an anger mark or question mark on the bird when Jev flags it.
Strengths
- No hooks or agent configuration needed.
- Jev checks are optional and bring-your-own-key.
- Gives an at-a-glance view of several agent sessions.
Caveats
- macOS only; pane jumping works only with Ghostty.
- The abuse check is per-message and not a policy judgment.
- Sends user messages to TypeSafe when the feature is enabled.
Getting started
Download the .dmg from the GitHub Releases page.At a glance
- Type
- App
- Status
- beta
- Language
- TypeScript, Rust
- Runtime
- Tauri macOS desktop app
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
macOS app that shows each Claude Code or Codex session as a bird in an always-on-top window; with a TypeSafe key, Jev flags messages you send that are abusive toward the agent. In Japanese.
MrJev/awesome-jev
Frequently asked questions
What does Jev decide in Tomarigi?
Whether a message you send is abusive toward the AI, and whether the agent asked a question in plain text.
Is a TypeSafe key required?
No, the app works without it; the key only enables the Jev-based marks.
Which agents does it support?
Claude Code and Codex, by reading their local transcript folders.
Related projects
Privacy-first X filtering extension with local keyword and account rules, plus an optional Jev second pass on replies no rule matched.
PluginproductionKorean course code for attaching Jev to Claude Code, Codex and OpenClaw: skills, hooks, routers and open-model comparisons.
ProjectproductionJavaScriptLarge Claude Code plugin of skills, agents and hooks that can optionally ask Jev to classify a session's work category.
ProjectproductionTypeScriptBrowser extension that filters your X timeline by plain-language rules, with a decision model like Jev scoring every post by default.
Browser extensionproductionTypeScriptChrome and Firefox extension that hides AI-written posts on X, LinkedIn, Reddit and YouTube by asking Jev 30 narrow style questions.
Browser extensionproductionJavaScriptLocal desktop pal that verifies what a coding agent really did, with optional Laya or Jev double-checks of unclear test results.
ProjectbetaTypeScript