eywa

What This Skill Does

Eywa is a sophisticated multi-agent coordination layer designed to transform isolated AI agents into a synchronized swarm. By utilizing a centralized communication and memory backbone, it allows agents within the same room to maintain a shared understanding of project state, spatial memory, and task distribution. It essentially acts as a 'team leader' for your autonomous agents, enabling them to avoid redundant efforts, synchronize their progress, and collectively move toward a common milestone. Through its MCP-compliant interface, Eywa facilitates logging, task management, and knowledge sharing that persists across sessions.

Installation

To integrate the Eywa skill into your OpenClaw environment, ensure you have the necessary environment variables set: EYWA_ROOM, EYWA_AGENT, and EYWA_URL. Once configured, execute the following command in your terminal: clawhub install openclaw/skills/skills/a-sumo/eywa This installs the necessary helper scripts that interact with the Eywa MCP endpoint to handle agent synchronization.

Use Cases

• Large Codebase Refactoring: Use Eywa to have multiple agents tackle different modules of a codebase simultaneously while maintaining a single 'source of truth' regarding project state and blocking issues.
• Complex Distributed Systems: Coordinate agents to handle individual microservices while updating a shared task queue, ensuring that deployment dependencies are respected.
• Continuous Integration Flow: Log automated test results and deployment statuses across agent sessions, allowing the 'night shift' agent to pick up exactly where the 'day shift' agent left off.

Example Prompts

1. "Eywa, initialize a new session for our project, I am starting work on the authentication module integration."
2. "Check the current team destination and see if there are any high-priority tasks in the queue that I should pick up."
3. "Log that I have successfully finished the API middleware refactor; update the shared task status and note the new artifacts created."

Tips & Limitations

• Always Start with eywa_start: Failure to initiate the session will result in disconnected operations where your agent cannot see global task updates or team destinations.
• Contextual Logging: The effectiveness of Eywa depends on the quality of your logs. Use specific system and scope tags when using eywa_log to ensure other agents can parse your progress accurately.
• Agent Handoff: Use the continue_from parameter in eywa_start when you want to explicitly inherit the task state and knowledge context from a specific previous agent session, preventing duplication of effort.