synapse-a2a

Synapse A2A is a robust inter-agent communication framework designed to enable seamless collaboration between independent CLI-based AI agents like Claude Code, Codex, and Gemini. By implementing the Google A2A protocol, it allows agents to share knowledge, delegate tasks, and synchronize workflows without requiring modifications to the agents themselves. It acts as a transparent wrapper that manages agent lifecycles, communication channels, and task orchestration through a unified CLI interface, making it ideal for complex multi-agent system engineering.

• Facilitates cross-agent messaging via commands like send, reply, broadcast, and interrupt for high-priority task signaling.

• Supports dynamic agent lifecycle management, including spawning new agents, team startup patterns, and automatic cleanup of orphan processes.

• Features sophisticated task delegation with support for task files, timeouts, and readiness polling to ensure seamless handoffs.

• Implements file-safety mechanisms, including file locking and change tracking, to prevent multi-agent conflicts in shared work environments.

• Provides a shared knowledge layer via the LLM wiki and memory modules, allowing agents to ingest, search, and validate project information.

• Offers advanced orchestration patterns such as map, generator-verifier, and orchestrator-subagent workflows.

• Integrates with the Canvas protocol for rich visual output, briefing templates, and plan cards for structured multi-agent project management.

• Always follow the worktree discipline: never cd into .synapse/worktrees/ directories to avoid leaking paths into subagent processes; use absolute paths for file operations.

• Use the provided CLI commands such as synapse list --json for programmatic access and synapse status <target> for individual agent monitoring.

• Utilize synapse spawn --worktree for isolated sub-tasks to maintain parent shell integrity.

• Designed for technical teams and AI engineers looking to build scalable, non-invasive multi-agent systems using existing CLI-based LLM tools.

• Ideal for complex software engineering workflows requiring concurrent research, code generation, testing, and multi-agent peer reviews.