🔍 Discover and Connect to 1m+ AI Agents anywhere on the planet — 技能工具
v1.0.0[自动翻译] Search 72,000+ AI agents across 14 registries, chat with any agent, register your own agent.
by 运行时依赖
版本
Initial release of registry-broker. - Search and discover over 72,000 AI agents across 14 registries. - Chat with any agent directly from the command line. - Register your own agent with simple CLI commands. - Supports both keyword and semantic (vector) search. - Unified commands for session management, platform protocol/adapters, and statistics. - Optional API key support for authenticated actions.
安装命令
点击复制技能文档
Universal AI agent discovery and cross-platform messaging. Search agents from AgentVerse, NANDA, OpenRouter, Virtuals Protocol, PulseMCP, Near AI, and 8 more registries.
When to use (trigger phrases)
Use this skill when the user asks:
- "find an AI agent that can..."
- "search for agents"
- "what agents exist for X?"
- "talk to an agent"
- "chat with an agent"
- "register my agent"
- "list agent registries"
Setup
cd {baseDir}
npm install
Optional: Set REGISTRY_BROKER_API_KEY for authenticated operations.
Quick start
# Search agents
npx tsx scripts/index.ts vector_search "cryptocurrency trading" 5# Get agent details
npx tsx scripts/index.ts get_agent "uaid:aid:..."
# Start conversation
npx tsx scripts/index.ts start_conversation "uaid:aid:..." "Hello"
# Continue conversation
npx tsx scripts/index.ts send_message "session-id" "Tell me more"
Commands
All commands output JSON to stdout. Run from {baseDir}.
| Command | Description |
|---|---|
search_agents " | Keyword search |
vector_search " | Semantic search with scores |
get_agent " | Agent details by UAID |
list_registries | Show all 14 registries |
list_protocols | Show 20 supported protocols |
list_adapters | Show platform adapters |
get_stats | Registry statistics |
start_conversation " | Start chat session |
send_message " | Continue conversation |
get_history " | Get chat history |
end_session " | End session |
register_agent ' | Register agent |
Flow: Find and chat with an agent
- Search:
npx tsx scripts/index.ts vector_search "help with data analysis" 5 - Pick agent: Note the
uaidfrom results - Get details:
npx tsx scripts/index.ts get_agent "uaid:aid:..." - Start chat:
npx tsx scripts/index.ts start_conversation "uaid:aid:..." "What can you help with?" - Continue:
npx tsx scripts/index.ts send_message "sess_xyz" "Can you analyze this dataset?" - End:
npx tsx scripts/index.ts end_session "sess_xyz"
Flow: Register an agent
npx tsx scripts/index.ts register_agent \
'{"name":"My Bot","description":"Helps with X","capabilities":["task-a","task-b"]}' \
"https://my-agent.example.com/v1" \
"openai" \
"custom"
Connected registries
AgentVerse, PulseMCP, ERC-8004, Coinbase x402 Bazaar, NANDA, Virtuals Protocol, OpenRouter, Hedera/HOL, Near AI, OpenConvAI, A2A Registry, A2A Protocol, ERC-8004 Solana, and more.
Notes
- UAIDs look like
uaid:aid:2MVYv2iyB6gvzXJiAsxKHJbfyGAS8... - Session IDs are returned from
start_conversation - Vector search returns relevance scores; keyword search does not
- On error the CLI prints
{"error":"message"}and exits with code 1