Home

Connect all the agents!

Key Features¶

🔌 ACP Integration¶

First-class support for the Agent Client Protocol (ACP):

• Integrate directly into IDEs like Zed, VS Code, and others
• Wrap external agents (Claude Code, Goose, Codex, fast-agent) as nodes
• Unified node abstraction - ACP agents work like native agents
• Compose ACP agents into teams with native agents

📝 Easy Agent Configuration¶

AgentPool excels at static YAML-based agent configuration:

• Define agents with extreme detail in pure YAML (Pydantic-backed)
• Expansive JSON schema for IDE autocompletion and validation, backed by an extremely detailed schema.
• Multi-Agent setups with native as well as remote (ACP / AGUI) agents

🧩 Unified Node Architecture¶

Everything is a MessageNode - enabling seamless composition:

• Native agents with a large set of default tools
• ACP agents
• AG-UI agents
• Teams (parallel and sequential)
• Human-in-the-loop-agents
• All nodes share the same interface

Dependencies¶

License¶

MIT License - see LICENSE for details.

Quick Start¶

Basic Agent Configuration¶

# agents.yml
agents:
  assistant:
    name: "Technical Assistant"
    model: openai:gpt-4
    system_prompt: You are a helpful technical assistant.
    tools:
      - type: file_access

Python Usage¶

from agentpool import AgentPool

async def main():
    async with AgentPool("agents.yml") as pool:
        agent = pool.get_agent("assistant")
        response = await agent.run("What is Python?")
        print(response.data)

if __name__ == "__main__":
    import anyio
    anyio.run(main)