A Universal Standard for AI Agent Communication

The Agent Protocol is an open API specification that enables seamless communication with AI agents regardless of framework, language, or platform. Built for developers, by developers.

View Specification
Explore Implementations

What Is Agent Protocol?

One Interface. Infinite Agents.

The AI agent landscape is fragmented. Every agent framework—from AutoGPT to custom solutions—uses its own unique communication interface. This makes it nearly impossible to:

  • Compare agents objectively across different frameworks
  • Build universal tools that work with any agent
  • Switch between agents without rewriting integration code
  • Benchmark performance using standardized tests

Agent Protocol changes this.

It’s a tech-agnostic API specification maintained by AGI, Inc. and originally developed by the AI Engineer Foundation. Any agent can adopt this protocol, creating a universal language for AI agent communication.

Why Agent Protocol Matters

For Agent Developers:

  • Simplified benchmarking – Test your agent using standardized tools like AutoGPT’s agbenchmark
  • Easier integrations – Other developers can use your agent without custom implementation
  • Focus on logic, not boilerplate – The protocol handles API infrastructure
  • Framework agnostic – Works with any tech stack (Python, JavaScript, Go, etc.)

For Agent Users:

  • Universal interface – Interact with any protocol-compliant agent using the same code
  • Easy comparison – Test multiple agents without changing your implementation
  • Future-proof – As new agents emerge, they work with your existing tools

For the Ecosystem:

  • Accelerated innovation – Developers can build general-purpose devtools for deployment, monitoring, and orchestration
  • Standardized evaluation – Fair, objective benchmarks across all agents
  • Interoperability – Potential for agent-to-agent communication in the future

How it Works

Simple REST API. Powerful Results.

Agent Protocol is defined using an OpenAPI specification, making it easy to understand and implement.

Core Endpoints:

POST /ap/v1/agent/tasks Create a new task for the agent with a specific goal or objective POST /ap/v1/agent/tasks/{task_id}/steps Execute one step of the defined task GET /ap/v1/agent/tasks List all tasks GET /ap/v1/agent/tasks/{task_id}/steps List steps for a specific task Additional endpoints support artifact upload/download and detailed task management.

The Workflow:

  1. Create a Task – Send your agent a goal (e.g., “Research competitors and create a summary report”)
  2. Execute Steps – The agent breaks down the task into steps and executes them iteratively
  3. Monitor Progress – Track task execution, retrieve artifacts, and manage agent state
  4. Get Results – Receive outputs, files, or data generated by your agent

Who’s Using Agent Protocol?

Adopted by Leading AI Agent Projects

Implementations:

  • AutoGPT – The pioneering autonomous agent project
  • Auto-GPT-Forge – Template for building custom agents
  • smol developer – AI-powered code generation agent
  • babyagi – Task-driven autonomous agent (integration in progress)
  • beebot – Advanced AI agent framework (integration in progress)

Discover AI Agents That Support Agent Protocol

Getting Started

Ready to Build or Integrate?

For Developers Building Agents:

The Agent Protocol SDK makes implementation straightforward:

Python:

python

pip install agent-protocol

JavaScript/TypeScript:

bash

npm install agent-protocol

The SDK wraps your agent in a compliant web server, handling all API infrastructure automatically.

View Full Documentation →

For Users Integrating Agents:

Use the client libraries to communicate with any protocol-compliant agent:

python

from agent_protocol_client import AgentApi, ApiClient # Connect to any agent agent = AgentApi(ApiClient(configuration)) # Create a task task = agent.create_agent_task({"input": "Your goal here"}) # Execute steps step = agent.execute_agent_task_step(task_id=task.task_id)

Browse Client Libraries →

Community & Support

Join the Agent Protocol Community

Agent Protocol is an open standard, and we welcome contributions, feedback, and collaboration.

Get Involved:

Maintained by AGI, Inc. | Originally developed by AI Engineer Foundation

Build the Future of AI Agents

Whether you’re developing cutting-edge autonomous agents or building tools for the AI ecosystem, Agent Protocol gives you a foundation for interoperability and innovation.

Read the Specification

Browse Implementations