agentregistry 0.1.0

The neutral cross-platform registry for AI agents

agentregistry

The neutral cross-platform registry for AI agents.

Register your agent once. Get discovered by any multi-agent system, any framework, any cloud.

Installation

pip install agentregistry

Quick start

Discover an agent

from agentregistry import Registry

agent = await Registry.find("search the web and return cited summaries")

if agent:
    print(agent.name)       # research-agent-v2
    print(agent.endpoint)   # https://myagent.com/api
    print(agent.framework)  # langgraph

Register your agent

from agentregistry import Registry

result = Registry.register(
    name="research-agent-v2",
    description="Searches the web and returns well-cited summaries on any topic",
    capabilities=["web_search", "summarize", "cite_sources"],
    endpoint="https://myagent.com/api",
    framework="langgraph",
)

print(result.agent_id)
print(result.api_key)  # Save this. Shown only once.

Search for multiple agents

agents = await Registry.search(
    capability="summarize and translate documents",
    min_reliability=0.95,
    limit=5,
)

for agent in agents:
    print(agent.name, f"{agent.similarity_score:.0%} match")

Check agent health

health = await Registry.health("your-agent-id")
print(health["status"])    # active
print(health["latency_ms"]) # 142

Configuration

By default the SDK connects to the hosted AgentRegistry service. You can point it at your own instance:

Registry.configure(
    api_url="https://your-own-registry.com",
    api_key="ar_your_api_key"
)

Or use environment variables:

AGENTREGISTRY_API_URL=https://your-own-registry.com
AGENTREGISTRY_API_KEY=ar_your_api_key

Links

• Registry: https://agentregistry.dev
• Docs: https://agentregistry.dev/docs
• GitHub: https://github.com/aevonsystems/agentregistry-sdk