Manage AI agents, not terminals.
Organize work across multiple projects as epics and tasks. Run AI agents in parallel on your machine — you review every decision before it merges.
How It Works
From backlog to pull request — you review, you merge.
Create Tasks
Organize features, bugs, and tech debt as epics. Use the kanban board or let AI decompose your brief.
Hit Play
Djinn spawns AI agents in isolated git worktrees. Each task gets its own sandbox. Dependencies respected.
Review & Merge
AI reviewers check each task against your acceptance criteria. You review the finished work and decide when to merge.
Parallel Execution
Run multiple AI agents in parallel, each in its own git worktree. Manage tasks on a kanban board instead of switching between terminal windows.
Local-First
Everything runs on your machine. No cloud. No data leaving your infrastructure. Enterprise-ready from day one.
Mix & Match Models
Use Claude for coding, GPT for reviews, Gemini for research — all at the same time. Set which models handle which tasks and at what priority. No manual switching.
Multi-Project
Microservices, monorepos, multiple projects — Djinn manages them all in parallel. Each repo has its own task database and knowledge base. One app to direct everything.
Any LLM Provider
Works with any provider supported by OpenCode — Claude, GPT, Gemini, and more. Use your existing subscription plans or API keys. No vendor lock-in.
You Define the Work
Most agents are stateless — they forget everything when you close the chat. Djinn organizes work as Epics, Stories, and Tasks that persist across weeks. You set the priorities, define the dependencies, and control what gets built and when.
You Control What the AI Knows
Decisions, patterns, and architectural rules live in a human-readable knowledge base — markdown files you can read, edit, and version control. You decide what context agents get. Your 100th task is informed by every decision you've documented.
Get Started
Free during beta. Bring your own API keys. You stay in control.