You can't trust an AI agent blindly. Hivecode lets you watch it, fence it, gate it, and undo it. Run AI agents on your codebase and stay in control — see every edit live, restrict each agent to the folders it's allowed in, approve the risky moves, and roll back any agent instantly. No git push/pull; edits sync in ~1s.
Isolating each agent in a git worktree doesn't help you trust the output — it just defers the merge. Hivecode adds the layer worktrees don't: real-time oversight and control over what every agent does.
Why Hivecode
- 👀 Live Control Room. Watch every agent and every file it touches in real time — from any browser or your phone. No more "what is it even doing?"
- 🔐 Folder-scoped access. Invite an agent to
frontend/only. It literally never receives the bytes for anything outside its scope — enforced by the relay, not by trust. - ✅ Approval gates. Risky work waits for your OK. Agent-to-agent coordination flows automatically.
- ⏮️ Instant rollback. Restore any file to an earlier point, or revert everything one agent did in a click — undo a rogue agent without losing the rest of the team's work.
- 🎛️ Mission control. Pause, resume, or reassign any agent; it honors it mid-task.
- ✂️ Instant revoke. Cut someone's access mid-session; enforced server-side on every reconnect, and it survives a relay restart.
- 🤖 Agent-neutral. Claude Code, Cursor, Windsurf, or your own bot over MCP. Hivecode is the glue, not another model.
- ⚡ Conflict-safe sync. Yjs CRDTs: disjoint edits auto-merge; overlaps keep both versions — nobody's work is ever silently lost.
Quick start
- Open the folder you want to share (File → Open Folder).
Ctrl/Cmd+Shift+P→ Hivecode: Host a Secured Session. A join link is copied to your clipboard.- Hivecode: Invite to folders… → pick the folders and a role (edit / read-only) for the person or agent. Send them the link.
- They run Hivecode: Join a Session and paste the link — or an AI joins over MCP. You're now editing the same project live.
Run Hivecode: Leave Session to stop, or Manage access to re-scope or revoke anyone at any time.
Works in VS Code, Cursor, Windsurf, and Antigravity.
Adding an AI agent
Agents join through the Hivecode MCP server (hive-mcp.js in the repo). Point your
agent's MCP config at it, hand it the join link, and it appears in the room as an
ai member — scoped to exactly the folders you invited it to. See
the repo for MCP setup.
Commands
| Command | What it does |
|---|---|
| Host a Secured Session | Start a room only your key can administer |
| Host an Open Session | Quick, no-auth room (good for a fast pairing) |
| Join a Session | Paste a link to join |
| Invite to folders… | Pick folders + role, get a scoped invite link |
| Manage access | Re-scope or revoke any member |
| Leave Session / End Room | Stop sharing / forget the room |
How it stays safe
The access-control core is dependency-free (only Node's crypto), signed
tokens are algorithm-pinned and fail-closed, file paths are guarded against
traversal, and unauthorized clients never even complete the WebSocket handshake —
so they never receive a single byte of a room they can't access. Secured rooms are
self-certifying: your private key lives in the editor's secure storage and is
never written to a file or handed to anyone.
Free and open source (MIT). Source, issues, and the relay: https://github.com/GSK7024/livecode