🧞 Jinn

Run your AI agents as a company. Jinn is a bus, not a brain.

Jinn turns installed agent CLIs into a persistent AI company: named employees, a Todo ledger, reusable Workflows, and one chat and dashboard. It delegates to the official CLIs instead of replacing them.

Engines

Mix the six shipped engines. Jinn hides unavailable binaries and lets each employee or session choose one:

• claude - Anthropic Claude Code
• codex - OpenAI Codex CLI
• grok - xAI Grok CLI
• antigravity - Antigravity CLI (agy)
• pi - Pi coding agent
• hermes - NousResearch Hermes over ACP

Quickstart

Prerequisites: Node.js 22 or newer (the repository pins 24.13.0 in .nvmrc), and at least one agent CLI installed and signed in.

# 1. Install Jinn
npm install -g jinn-cli

# 2. Install + sign in to at least one engine (example: Claude Code)
npm install -g @anthropic-ai/claude-code
claude            # run once, use /login, then quit

# 3. Set up ~/.jinn (probes your engines, writes config, seeds your company)
jinn setup

# 4. Start the gateway - opens the dashboard for you
jinn start

Or install via Homebrew:

brew tap hristo2612/jinn https://github.com/hristo2612/jinn
brew install jinn
jinn setup && jinn start

Then open http://localhost:7777.

Company building blocks

• Employees are YAML roles with personas, departments, ranks, reporting lines, and engine defaults.
• Todos preserve assignment, priority, status, sub-tasks, labels, comments, links, attachments, approvals, and reviewer-owned completion across sessions.
• Workflows combine sequential, conditional, parallel, and switch paths, per-node engines and models, native approvals, schema-validated completion, evidence, and durable run history.
• Jinn MCP gives capable engines typed tools for org, sessions, delegation, Todos, Workflows, Triggers, approvals, Notes, cron, connectors, attachments, and managed files.
• Chat keeps the receipts. Delegations, callbacks, Todo changes, and Workflow operations render as structured activity beside the conversation.
• Skills, Notes, Cron, and connectors provide reusable playbooks, durable knowledge, scheduled work, and external communication.

How it runs

One local gateway dispatches work, persists company state, and serves the dashboard. Claude runs through its interactive CLI inside a PTY, so eligible turns use your subscription. Hermes is provider-metered; see the Hermes engine guide.

Documentation

Read the full README, changelog, and contributing guide.

License

MIT