See everything
Live trace of every prompt, response, and tool call. Watch file edits, bash commands, and API calls as they happen.

You run aimux and instantly see every AI agent running on your machine: Claude, Codex, Gemini. For each one, you get a live trace of what it’s doing, turn by turn. Then you judge it.
See everything
Live trace of every prompt, response, and tool call. Watch file edits, bash commands, and API calls as they happen.
Judge quality
Label turns GOOD, BAD, or WASTE. Add notes. Build eval datasets from real production sessions.
Export anywhere
OTEL to MLflow or Jaeger. JSONL to disk. Your annotations become feedback assessments for model evaluation.
Zero setup
Auto-discovers running agents. No config, no instrumentation, no SDK. Just run aimux and it finds them.
Run aimux for a keyboard-driven TUI with split-pane trace and live session embedding. Or run aimux web for a browser dashboard with card grid, filtering, dark/light themes, and a plugin system.


| Feature | Description |
|---|---|
| Cost tracking | Per-turn pricing for Claude, Codex, Gemini models with project-level breakdowns |
| Session history | Browse, search, and revisit past sessions across all providers |
| Launch agents | Spawn agents in tmux (persistent) or direct mode from the dashboard |
| Google Tasks | Launch agents with task context, track completion automatically |
| Cross-session search | Find which agent edited a specific file across all sessions |
| Notifications | Terminal bell + macOS desktop alerts on permission prompts, errors, completion |
| Plugins | Custom dashboard tabs with metrics, tables, and charts (e.g., Skill Dashboard) |
| Dark/light themes | Toggle in the web dashboard, persisted to localStorage |
| Kubernetes | Run agents on K8s pods, control from your laptop via Redis coordination |
brew install zanetworker/aimux/aimuxaimux # TUI dashboardaimux web # web dashboard