all projects

Swarm MCP Lab

A coordination layer for running multiple AI agents in parallel on a single machine — with a native macOS visualization that lets an operator actually see what their swarm is doing in real time.

AI/MLAgentsmacOSR&D
Swarm MCP Lab hero

Vision

Most "multi-agent" systems today are either single agents pretending, or expensive cloud orchestrators. Swarm is the missing primitive underneath: a purpose-built local coordination layer that lets independent agent sessions discover each other, hand off work, and share live context without a central orchestrator. The companion native macOS visualization renders the active swarm as a live, animated graph — so the human in the loop has real situational awareness instead of grepping logs across five terminals. Architectural details intentionally not detailed here.

Built with

TypeScriptRustNative macOSMulti-Agent Systems

Gallery

Live multi-agent graph: animated nodes, message pulses on edges, embedded terminal panels per agent.
Spawning and managing agents from the dashboard with real-time visibility into shared state.