What is CLI-JAW?

CLI-JAW is an open-source platform that unifies every AI coding CLI you already use — Claude, Codex, Gemini CLI, OpenCode, Copilot CLI — into one assistant with one memory and one dashboard. Your main CLI calls the others as "employees" (sub-agents), so you stop copy-pasting between heavy apps and start giving orders from a single place.

It routes through the subscriptions you already pay for. No per-token API billing. No rate limits on top of what you already have.

One developer is managing 100k+ lines of code with this thing. It's basically a super-app: multi-instance manager + mini-Obsidian notes + Kanban board + browser automation, running locally on your machine.

What's New in 2.0.0

This is the general availability release. Here's what landed:

Plus: improved PABCD orchestration, 230+ skills (32 active, 194 reference), SQLite FTS5 memory search, and Computer Use desktop automation.

Two-Line Install

npm install -g cli-jaw
jaw dashboard

That's it. Open http://localhost:3457. Requires Node.js 22+.

jaw serve also works if you prefer the classic server mode.

Auto-start on boot: jaw service install (auto-detects systemd / launchd / Docker).

Platform Setup

curl -fsSL https://raw.githubusercontent.com/lidge-jun/cli-jaw/master/scripts/install.sh | bash

jaw dashboard

wsl --install

curl -fsSL https://raw.githubusercontent.com/lidge-jun/cli-jaw/master/scripts/install-wsl.sh | bash

source ~/.bashrc
jaw dashboard

npm install -g cli-jaw
jaw dashboard

Authenticate

You only need one. Pick whichever subscription you already have:

# Free options
copilot login        # GitHub Copilot (free tier available)
opencode             # OpenCode — free models available
Paid (monthly subscription you already pay for)
claude auth login    # Anthropic Claude Max
codex login          # OpenAI ChatGPT Pro
gemini               # Google Gemini Advanced

Check everything at once: claude auth status && jaw doctor

🦈 CLI-JAW Doctor — 12 checks

 ✅ Node.js        v22.15.0
 ✅ Claude CLI      installed
 ✅ Codex CLI       installed
 ⚠️ Gemini CLI      not found (optional)
 ✅ OpenCode CLI    installed
 ✅ Copilot CLI     installed
 ✅ Database        jaw.db OK
 ✅ Skills          32 active, 194 reference
 ✅ MCP             3 servers configured
 ✅ Memory          structured/ exists
 ✅ Server          port 3457 available

The Dashboard

The dashboard is the main way to use CLI-JAW 2.0. It's a local web app that acts as command center for all your AI instances.

Instance Management

The navigator panel shows every JAW instance — running, stopped, or offline. For each one you can:

• Start / Stop / Restart with one click
• See the active CLI engine, model, port, and working directory
• Use perm mode for persistent instances that survive reboots
• Launch new instances pointed at different project directories

Live Preview

Select any running instance and its Web UI embeds directly in the dashboard via a preview proxy. Chat with your AI, see streaming responses, drag-and-drop files — all without leaving the manager.

Notes

A markdown vault lives inside the dashboard. Think mini-Obsidian:

• Folder tree with create/rename/move/delete
• Raw, split, and preview editing modes
• Dirty-state markers so you don't lose unsaved work
• KaTeX math rendering, Mermaid diagrams, syntax-highlighted code blocks

Kanban Board

Drag instance cards into lanes (Backlog / In Progress / Review / Done). Each card links to the live instance and has an editor for notes. Good for tracking what each AI session is working on when you're running multiple instances.

Settings Inspector

View and edit runtime settings for any selected instance: active CLI, model, reasoning effort, permission mode, working directory, employees, skills, MCP servers.

How the Employee System Works

This is the core idea: your main CLI calls other CLIs as workers.

You talk to one AI (the "Boss"). When it needs specialized work, it dispatches tasks to employees — each running their own CLI with their own model:

You: "Fix the frontend styling and update the API endpoint"
Boss (Claude) thinks...
├── Dispatches to Frontend employee (OpenCode) → "Fix the CSS grid layout in dashboard.tsx"
├── Dispatches to Backend employee (Codex)     → "Update /api/users to return pagination metadata"
└── Synthesizes both results for you

# Under the hood, it's one command:
jaw dispatch --agent "Frontend" --task "Fix the CSS grid layout in dashboard.tsx"

Employees are other AI CLIs configured in your settings. Each has its own session, its own model, its own context. The Boss reviews their output before presenting it to you.

Employees vs. Sub-agents

These are different things:

Use employees for "Frontend does CSS, Backend does API." Use sub-agents for "read these 5 files in parallel before deciding."

Five AI Engines

No per-token API billing. Route through subscriptions you already pay for.

Fallback chain: if one engine is rate-limited, the next picks up. Configure with /fallback [cli1 cli2...].

OpenCode wildcard: connect any model endpoint — OpenRouter, local LLMs, any OpenAI-compatible API.

Switch engines live: /cli codex. Switch models: /model gpt-5.5. Works from Web, Terminal, Telegram, or Discord.

PABCD Orchestration

For complex tasks, CLI-JAW uses a 5-phase workflow. You approve every transition.

P (Plan) → A (Audit) → B (Build) → C (Check) → D (Done) → IDLE
   ⛔          ⛔          ⛔         auto        auto

State is DB-persisted and survives restarts. Workers cannot modify files — only verify. Activate with jaw orchestrate or /pabcd from any interface.

Memory

Three layers, each covering a different recall horizon.

All three layers feed into the system prompt automatically. Memory is searchable:

jaw memory search "how did we set up the API auth?"

Skills

230+ skills covering dev workflows, office documents, automation, and media.

32 skills are always active. 194+ reference skills load on demand.

jaw skill install <name>    # activate a reference skill
jaw skill list              # see what's available

Browser & Desktop Automation

Computer Use lets you control any macOS app — Finder, Safari, System Settings, Xcode — through natural language. Point it at your localhost dev server in Safari and you get a full visual testing loop.

Messaging

Telegram

📱 Telegram ←→ 🦈 CLI-JAW ←→ 🤖 AI Engines

Text chat, voice messages (auto-transcribed via multi-provider STT), file/photo upload, slash commands (/cli, /model, /status), scheduled task delivery.

Discord

Same capabilities as Telegram — text, files, commands. Channel and thread routing with forwarder for agent result broadcast. Setup via Web UI settings.

Voice & STT

Voice input works on Web (mic button), Telegram (voice messages), and Discord. Providers: OpenAI-compatible, Google Vertex AI, or any custom endpoint.

MCP

Model Context Protocol lets AI agents use external tools. CLI-JAW manages MCP config for all five engines from one file.

jaw mcp install @anthropic/context7
# → syncs to Claude, Codex, Gemini, OpenCode, Copilot config files simultaneously

No more editing five different JSON files. Install once, every engine gets it.

jaw mcp sync       # re-sync after manual edits

CLI Commands

# Core
jaw dashboard                     # launch manager dashboard
jaw serve                         # start server (http://localhost:3457)
jaw chat                          # terminal TUI
jaw doctor                        # 12-point diagnostics
Instances
jaw clone ~/project               # clone instance to new directory
jaw --home ~/project serve --port 3458  # run second instance
jaw service install               # auto-start on boot
AI & Orchestration
jaw dispatch --agent "Backend" --task "..."  # dispatch employee
jaw orchestrate                   # enter PABCD workflow
Skills & MCP
jaw skill install <name>          # activate a skill
jaw skill list                    # list available skills
jaw mcp install <package>         # install MCP → syncs to 5 engines
jaw mcp sync                      # re-sync MCP configs
Memory
jaw memory search <query>         # search across all memory layers
jaw memory save <file> <content>  # save to structured memory
Browser
jaw browser start                 # launch Chrome (CDP)
jaw browser snapshot              # capture page state
jaw browser vision-click "Login"  # AI-powered click
Maintenance
jaw reset                         # full reset

Multi-Instance & Docker

Multi-Instance

Run isolated instances with separate settings, memory, and database:

jaw clone ~/my-project
jaw --home ~/my-project serve --port 3458

Each instance is fully independent — different working directory, different memory, different MCP config. The manager dashboard sees them all.

Docker

docker compose up -d       # → http://localhost:3457

Non-root jaw user, Chromium sandbox enabled. Data persists in jaw-data named volume.

# Dev build
docker build -f Dockerfile.dev -t cli-jaw:dev .
docker run -d -p 3457:3457 --env-file .env cli-jaw:dev

# Pin version
docker build --build-arg CLI_JAW_VERSION=2.0.0 -t cli-jaw:2.0.0 .

# If Chromium sandbox fails
docker run -e CHROME_NO_SANDBOX=1 -p 3457:3457 cli-jaw

Development

npm run build          # tsc → dist/
npm run dev            # tsx server.ts (hot-reload)
npm test               # native Node.js test runner

Architecture details: ARCHITECTURE.md · Test coverage: TESTS.md · Internal docs: devlog/structure/

How It Compares

Troubleshooting

Contributing

1. Fork and branch from master
2. npm run build && npm test
3. Submit a PR

Bug reports and feature ideas: Open an issue