femto/minion-agent

A simple agent framework that's capable of browser use + mcp + auto instrument + plan + deep research + more

What it solves

Minion Agent provides a unified framework for building AI agents that can perform complex tasks requiring browser automation, code execution, and deep research. It simplifies the process of integrating multiple agent frameworks (like smolagents, LangChain, and OpenAI Assistants) into a single interface, allowing developers to switch between different underlying engines based on the task requirements.

How it works

The project uses a configuration-based approach (AgentConfig) to define an agent's identity, tools, and the underlying framework. It supports multiple backends, with the EXTERNAL_MINION_AGENT framework serving as the default for high-performance tasks. The system integrates with the Model Context Protocol (MCP) to allow agents to connect to external tools via standard commands or Server-Sent Events (SSE).

Who it’s for

It is designed for developers building autonomous agents that need to interact with the web, execute code locally, or conduct extensive information gathering and research.

Highlights

  • Multi-Framework Support: Compatible with EXTERNAL_MINION_AGENT, smolagents, LangChain, OPENAI, BROWSER_USE, and DEEP_RESEARCH.
  • Browser Automation: Built-in capabilities for web interaction and control.
  • Code Execution: Advanced code generation and execution capabilities for solving technical tasks.
  • MCP Integration: Full support for the Model Context Protocol to easily add external tools.
  • Planning Capabilities: Supports automatic planning intervals to re-evaluate tasks every few steps (via smolagents).

Related

  • Project
    owainlewis/blueprint

    A collection of portable skill files that guide AI coding agents through professional software engineering practices like design, planning, and review.

  • Project
    InditexTech/mcp-teams-server

    An MCP server that integrates Microsoft Teams with LLMs, allowing AI agents to read messages, create threads, and mention members.

  • Project
    skygazer42/GustoBot

    A multi-agent AI customer service system for Chinese recipes that combines GraphRAG, Text2SQL, and multimodal capabilities to provide deep culinary knowledge.

  • Project
    stardomains3/oxproxion

    An Android chat application that provides a unified interface for interacting with cloud-based LLMs via OpenRouter and locally hosted models via Ollama, LM Studio, and others.