Skip to content
Get packs

MetaGPT

A framework where several LLM agents play software-company roles to build software from one sentence.

GitHub Download v0.8.1
Stars
71.1k
Forks
9.1k
Licence
MIT
Language
Python
Last update
10 months ago
Type
Framework

What it is

MetaGPT is a Python framework that assigns different roles, such as product manager, architect, project manager, and engineer, to LLM-based agents. You give it a one-line requirement and the agents work through a software-company style process to produce artifacts like user stories, requirements, data structures, APIs, and code. It is used by people who want to experiment with multi-agent collaboration on complex tasks.

  • Generate a small project from a one-line request, such as the README's example of creating a 2048 game
  • Run a data analysis task with the Data Interpreter component on a dataset and get a plot
  • Build your own agents using the agent and multi-agent tutorials in the docs
  • Explore the example use cases listed in the docs, such as Debate, Researcher, and Receipt Assistant

Great for

  • Developers experimenting with multi-agent LLM workflows
  • Researchers interested in agent collaboration using standard operating procedures
  • Students exploring how role-based AI teams are built
  • People who want a starting-point project to generate small software prototypes

Not ideal for

  • Users who need reliable, production-ready code without reviewing output, since the README does not make such guarantees
  • People without an LLM API key or local model setup, since configuration is required before use
  • Users who want a no-code, point-and-click tool; the README centers on CLI and Python usage
Step-by-step guide: how to use MetaGPTInstall, first run, examples, tips and FAQs
MetaGPT alternativesMetaGPT vs OpenHandsCrewAI vs MetaGPTMetaGPT vs OpenManusAgent Lightning vs MetaGPTEdict vs MetaGPT

How to get it

  1. 1

    Install with a package manager

    You'll need: Python 3.9 or later, but less than 3.12 · Node.js · pnpm · An LLM API key or compatible local model endpoint

    Install from PyPI

    pip install --upgrade metagpt

    Create initial config file

    metagpt --init-config
  2. 2

    Run it

    • Install Python 3.9 to 3.11, Node.js, and pnpm
    • Run pip install --upgrade metagpt
    • Run metagpt --init-config and edit ~/.metagpt/config2.yaml with your LLM settings
    • Run metagpt "Create a 2048 game" and check the generated repo in ./workspace
  3. 3

    Get the most out of it

    • Use the Data Interpreter for data analysis tasks, as shown in the README example
    • Follow the Agent 101 and MultiAgent 101 tutorials to build custom roles
    • Point the config file at other providers such as Azure, Ollama, or Groq, as the README notes
    • Check the examples directory for research work such as SPO and AOT referenced in the news section

Downloads

Source code (.zip)main branch · straight from GitHubRelease v0.8.1Release notes and sourceOfficial website & docsatoms.dev/

Strengths

  • Clear, documented multi-role design built around a software-company process
  • Supports several LLM providers through a single YAML configuration file
  • Open source under the MIT license with an active project and a large star count

Trade-offs

  • Output quality depends on the underlying LLM and API costs are paid by the user
  • The README requires a specific Python version range and Node.js with pnpm, which adds setup steps
  • Documentation is spread across an external site, so some details are not in the README itself

How it compares

ProjectStarsTypeLevelUpdatedBest for
MetaGPT (this)71.1kFrameworkintermediate10 months agoDevelopers experimenting with multi-agent LLM workflows
crewAI60.4kLibraryintermediate4 days agoPython developers prototyping multi-agent workflows
OpenHands92.5kAppintermediate4 days agoDevelopers who want coding agents running on a server while their laptop is off
deer-flow84.9kAI model / agentadvanced4 days agoDevelopers who want to experiment with multi-agent workflows on their own infrastructure
botpress15kDev toolintermediate6 days agoDevelopers who want to write integrations and bots as TypeScript code using the Botpress CLI and SDK
deep-research19.9kAI model / agentintermediate7 months agoDevelopers who want a readable example of a deep research agent to learn from or extend

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.