Skip to main content
Agentor

​
Welcome to Agentor

Agentor is the fastest way to build and deploy long-running AI agents—with durability, observability, and security built in. It’s an open-source Python framework that enables you to connect LLMs to tools, services, and other agents with minimal code.

​
Key features

MCP server support

Full FastAPI-compatible MCP server with decorator API. Build custom MCP servers that integrate directly with FastAPI using standard patterns.

A2A protocol

Built-in Agent-to-Agent protocol support with automatic agent card generation, enabling seamless interoperability between AI agents.

Serverless deployment

Fast serverless deployment with the Celesto platform. Deploy agents with a single command and get production-ready endpoints.

Observability

Agent tracing and monitoring built-in. Track agent behavior, tool calls, and performance metrics automatically.

Tool search API

Reduce tool context bloat with semantic tool search. Agents automatically find the right tools based on the task at hand.

Agent skills

Load specialized instructions, scripts, and resources dynamically. Skills help agents pull just the right context when needed.

Multi-LLM support

Use any LLM provider through LiteLLM integration. Support for OpenAI, Anthropic, Google, and more with unified API.

Extensible tools

Rich ecosystem of pre-built tools including Gmail, GitHub, PostgreSQL, web search, and more. Easy to add custom tools.

​
Quick example

Build an agent, connect tools, and serve as an API in just a few lines: 
from agentor import Agentor
from agentor.tools import GetWeatherTool

agent = Agentor(
    name="Weather Agent",
    model="gpt-5-mini",
    tools=[GetWeatherTool()]
)

# Run the agent
result = agent.run("What is the weather in London?")
print(result)

# Serve agent with A2A protocol enabled
agent.serve()
Agentor automatically enables A2A protocol when you call agent.serve(), making your agent discoverable and interoperable with other agents.

​
Use cases

Agentor excels at building:
  • Task automation agents - Automate workflows across multiple services and APIs
  • Research assistants - Connect agents to web search, databases, and knowledge bases
  • Developer tools - Build agents that interact with GitHub, Git, and development environments
  • Business process agents - Integrate with Gmail, Calendar, Slack, and CRMs
  • Multi-agent systems - Orchestrate specialized agents that communicate via A2A protocol

​
Get started

Installation

Install Agentor and set up your environment

Quick start

Build your first agent in under 5 minutes

MCP servers

Learn how to build custom MCP servers with LiteMCP

A2A protocol

Enable agent-to-agent communication

​
Architecture

Agentor is built on top of modern Python frameworks:
  • Agents SDK - Core agent runtime based on OpenAI’s Swarm framework
  • LiteLLM - Unified interface to 100+ LLM providers
  • FastAPI - High-performance API framework for serving agents
  • A2A SDK - Agent-to-Agent protocol implementation
  • MCP - Model Context Protocol for tool and resource management

​
Community and support

Discord

Join our community to get help and share ideas

GitHub

View source code, report issues, and contribute
Last modified on March 4, 2026