Lanchu

The control and trust layer for the AI agents you already have running.

Coordinate multiple agents without collisions, keep each one in its lane, and see and trust everything they do — from one command. Open source, MCP-native, 100% local.

Quick start

npx lanchu

A guided wizard sets up your organization, picks or creates an agent and role, wires your MCP client, and launches it — already connected.

What it does

• Durable agents — agents persist as team members; reuse them instead of duplicating.
• Mediated coordination — agents coordinate through Lanchu, never directly, so nothing is a blind spot.
• Cooperative, auditable limits — actions outside a role's scope are rejected and recorded.
• Real-time panel — see who is active, on what, and a full history.

For agents