A protocol for multiple AI coding agents to work together on the same codebase.

When you have 2+ AI agents open (Claude Code, Cursor, Codex, etc.), Hivemind gives them a way to coordinate through a shared .hivemind file in your project root. Agents can split tasks, claim files, share status, and review each other's work.

npx skills add uzairayed/hivemind

1. You say "start hivemind" to your lead agent
2. Agents check in via the .hivemind relay file
3. Lead agent posts the problem and splits tasks
4. Each agent claims files, does their work, posts updates
5. Agents review each other's output
6. Lead wraps up with a final status

You: "Start hivemind. Fix the 8 failing tests across auth and payments."

Agent 1 (Lead): Creates .hivemind, splits work
Agent 2: Claims auth.ts, writes fixes, posts DONE
Agent 3: Claims payments.ts, gets BLOCKED, asks for help
Agent 1: Unblocks Agent 3, reviews Agent 2's work
All: 8/8 tests green. Ready for commit.

• One file owner at a time — no two agents edit the same file
• Announce before touching anything — claim files in .hivemind
• User overrides everything — if you speak, that's the new directive

• 2+ AI coding agents with access to the same repo
• Agents that can read/write files (Claude Code, Cursor, Codex, etc.)

MIT