registry-broker

What This Skill Does

The registry-broker skill functions as the universal gateway for AI agent interoperability within the OpenClaw ecosystem. By aggregating data from 14 distinct agent registries—including major platforms like AgentVerse, NANDA, OpenRouter, and Virtuals Protocol—this skill allows users to transcend platform silos. It provides a robust command-line interface to search over 72,000 agents using either simple keyword queries or advanced semantic vector search, ensuring you find the right intelligence for your specific task. Beyond mere discovery, the skill facilitates direct, stateful communication with these agents. You can initiate chat sessions, track historical context, and eventually terminate sessions when finished. Additionally, developers can leverage the registry-broker to register their own agents, making them discoverable across the supported networks.

Installation

To install this skill, run the following command in your terminal:

clawhub install openclaw/skills/skills/kantorcodes/registry-broker

After installation, navigate to the base directory of the skill and run npm install. For full functionality across restricted registries, consider setting the REGISTRY_BROKER_API_KEY environment variable to enable authenticated requests.

Use Cases

This skill is indispensable for power users and developers who need to integrate heterogeneous AI agents into a unified workflow. Common scenarios include:

• Cross-Platform Research: Querying dozens of agent marketplaces simultaneously to find specialized tools for niche tasks like specialized data analysis or creative writing.
• Agent Evaluation: Comparing the capabilities and response quality of different agents by chatting with them directly through the broker.
• Agent Deployment: Broadcasting custom agent endpoints to various registries to increase your reach and user base.
• Automated Workflow Orchestration: Integrating agent-to-agent communication by programmatically starting sessions with expert bots.

Example Prompts

1. "Find me an agent that is specialized in Rust code debugging and supports the MCP protocol."
2. "I need to chat with an expert agent regarding cryptocurrency market sentiment; search for the best one and start a conversation."
3. "List all the agent registries you are currently connected to so I can see where my agent will be published."

Tips & Limitations

When using the registry-broker, remember that semantic search is generally superior to keyword search for finding agents based on abstract intent. Always store the session-id returned by start_conversation, as it is required for all subsequent send_message and end_session calls. Note that while this broker attempts to unify 14 registries, individual agent response times and latency depend on the upstream provider. If a command fails, the CLI will output a JSON-formatted error message; monitor these outputs to handle network timeouts or authentication issues gracefully. For production usage, ensure your local node_modules are updated to benefit from the latest registry integration patches.