5 Engines · One Brain · Always On
Get your personal AI assistant in 2 lines
Pick whichever you have. CLI-JAW auto-detects and falls back gracefully. Even one engine is enough.
Engine down? → Next one picks up automatically. Zero downtime.
Not just another chatbot. A full-stack personal assistant with real capabilities.
Uses official CLIs only — no API key scraping, no reverse engineering, no ban risk.
DB-persisted FSM pipeline — Plan → Audit → Build → Check → Done. Workers are read-only. You approve every phase.
Browser automation, document generation, Telegram, memory, GitHub, Notion — ready out of the box.
429 rate-limited? Same engine retries once, then auto-falls back to the next available engine. Zero downtime.
Chat from your phone. Send voice, photos, files. Get AI responses, generated docs, and screenshots back.
Install MCP once → syncs to all 6 AI engines instantly. One config, six agents.
macOS, Linux, WSL, Docker. ENOENT-safe spawn, auto-detection, container isolation.
A DB-persisted finite state machine that enforces planning, auditing, and verification. No cowboy coding — every phase requires your approval.
Audit and verification workers cannot modify files. They inspect, report, and recommend — but never touch your code.
The AI cannot self-advance. You review and approve each transition. No runaway agents, no surprise changes.
State survives server restarts. CLI and Web UI share the same FSM — resume from any interface.
Independent subtasks run concurrently via Promise.all with automatic file-overlap
detection and sequential fallback.
If your OS has a CLI, CLI-JAW works there. No exceptions.
Run CLI-JAW in a Docker container for security isolation. AI agents cannot access host files. Non-root user, Chromium sandbox enabled by default.
Dockerfile Dockerfile.dev docker-compose.ymlWeb UI, Terminal TUI, or Telegram. Same assistant, same memory, same skills.
