agent-protocol

Stale
GitHub Python MIT

Description

Common interface for interacting with AI agents. The protocol is tech stack agnostic - you can use it with any framework for building agents.

Key Features

  • Universal agent interface — single common REST API specification for communicating with any AI agent
  • Tech stack agnostic — adopt the protocol with any framework (Python, JS, or custom) without restrictions
  • Minimal core design — starts with two essential routes: create task and execute step, keeping adoption simple
  • OpenAPI specification — protocol defined in standard OpenAPI format for automatic code generation
  • SDK and client libraries — official Python SDK for agents and client library for users to interact with agents
  • Adopted by major projects — Auto-GPT, babyagi, smol developer, and others have implemented the protocol

Use Cases

💡 Agent interoperability — enabling different AI agents to communicate and collaborate through a standard interface
💡 Benchmarking — providing a consistent API to evaluate and compare agent performance across frameworks
💡 Devtool development — building monitoring, deployment, and debugging tools that work with any agent implementation
💡 Agent marketplace — allowing users to swap between different agent backends with minimal code changes

Categories

Quick Start

pip install agent-protocol → from agent_protocol import Agent → agent = Agent(your_agent_function) → agent.start() → POST /ap/v1/agent/tasks to create task → POST /ap/v1/agent/tasks/{task_id}/steps to execute steps

Related Projects