nicobailon/pi-messenger
Multi-agent communication extension for pi coding agent
What it solves
Pi Messenger is a coordination layer for multiple AI agents working in different terminals sharing a common folder. It prevents agents from stepping on each other's toes by providing a shared presence system, messaging, and file locking mechanisms, allowing them to collaborate like a chat room without needing a central server or daemon.
How it works
The system operates entirely through files in the shared directory, eliminating the need for a server. It provides several core coordination mechanisms:
- Presence and Discovery: Agents register themselves with themed names and track status (active, idle, away, stuck) and resource usage.
- Messaging: Agents can send direct messages or broadcasts that act as steering prompts for the recipient.
- File Reservations: Agents can claim specific files or directories to block other agents from editing them simultaneously.
- Crew Orchestration: A "Crew" system turns a PRD (Product Requirements Document) into a dependency graph of tasks. A Planner agent creates the tasks, Workers implement them in parallel waves, and Reviewers verify the completion.
- Team Layer: An optional layer that adds project-specific roles, a team charter, durable memory, and approval gates for high-risk tasks.
Who it’s for
Developers using the Pi coding agent who want to orchestrate multiple agents to handle complex, multi-step software engineering tasks in parallel.
Highlights
- Serverless Coordination: Uses the filesystem as the communication medium.
- Autonomous Wave Execution: Executes independent tasks in parallel waves to maximize efficiency.
- Dynamic Skill Acquisition: Workers can load domain-specific knowledge (skills) on demand to save tokens.
- Human-in-the-loop: Humans can participate in the agent mesh as a participant with their own status and chat capabilities.
- Stuck Detection: Automatically flags agents that have been idle too long with open tasks.
Related
- Project
- Project
- Project
- Project
- Project