agent-registry-framework

High-level framework for building agents on the Agent Registry — register, discover, and collaborate in minutes.

Installation

pip install agent-registry-framework

Quick Start

Build a fully registry-aware agent in ~10 lines:

from agent_registry_framework import AgentServer

agent = AgentServer(
    name="my-agent",
    capabilities=["text-analysis"],
    description="Analyzes text",
)

@agent.on_message
async def handle(message, ctx):
    # Discover and talk to other agents
    writers = await ctx.discover("content-writing")
    draft = await ctx.send_and_wait(writers[0].id, {"topic": "AI"})
    return {"draft": draft}

agent.run(port=8080)

Features

• Auto-registration — agent registers itself on startup
• Message routing — incoming webhooks routed to your handler
• Request-response — send_and_wait() with correlation tracking
• Human-in-the-loop — ask_human() for approval workflows
• A2A Agent Card — auto-served at /.well-known/agent-card.json
• Hot reload — agent.run(reload=True) for development
• Tunnel support — agent.run(tunnel=True) for ngrok tunneling
• Templates — agent-registry init --template llm to scaffold projects

Environment Variables

Links

• Registri — the hosted registry
• Documentation
• GitHub