MetaGPT
Multi-agent framework that assigns GPT roles (PM, engineer, QA) to solve complex software tasks end-to-end.
MetaGPT
MetaGPT encodes software company SOP (standard operating procedures) into multi-agent collaboration. Given a one-line requirement, MetaGPT spawns Product Managers, Architects, Engineers, and QA Testers as separate agents — each with defined roles, memory, and tool access — to collaboratively produce PRDs, system designs, complete code, and unit tests.
Key Features
- Role-based agents — PM, Architect, Engineer, QA, and Data Analyst roles with structured handoffs
- Shared message pool — agents communicate via a structured blackboard rather than direct prompts
- Incremental development — supports iterative code generation, review, and debugging loops
- Tool integration — agents can browse the web, run code in a sandbox, and call external APIs
- Multi-model backend — works with OpenAI, Anthropic, local Ollama, and any LiteLLM-compatible provider
- Data Interpreter — specialized sub-agent for end-to-end data science pipelines
Quick Start
pip install metagpt
metagpt "Write a CLI tool to convert CSV files to JSON"
import asyncio
from metagpt.software_company import generate_repo
async def main():
await generate_repo("Create a REST API for a todo list app")
asyncio.run(main())
Install via ai-supply
npx ai-supply add metagpt-multi-agent-framework
Curated mirror of the open-source MetaGPT (MIT). Get it from the source.
Compromise signals — malicious or tampered code (leaked secrets, backdoors, a dropped executable) — reduce the score, and known dependency CVEs carry a bounded penalty (they warrant review but never QUARANTINE — update the dependency to clear). Other dangerous-by-capability traits are risk surface, expected for some capabilities. Every finding is mapped to its OWASP control below.
Findings mapped to the OWASP Top 10 for LLM Applications (2025) and the OWASP Machine Learning Security Top 10. Expand any flagged control for the exact findings — compromise reduces the score; expected/risk-surface do not, except a known CVE, which carries a small bounded penalty (high/critical → Review).
The same gate an agent runs before installing (POST /api/v1/trust/metagpt-multi-agent-framework/check). Click a policy:
Consume MetaGPT programmatically. Authenticate with an API key or session — see Authorize an agent.
# Agents: CHECK BEFORE YOU INSTALL (no auth) — score, grade, level, capability manifest
curl https://ai-supply.store/api/v1/trust/metagpt-multi-agent-framework
# Gate against your org policy (returns { pass, violations })
curl -X POST https://ai-supply.store/api/v1/trust/metagpt-multi-agent-framework/check \
-H "Content-Type: application/json" \
-d '{"minGrade":"B","denyPermissions":["shell"],"denyUnknownEgress":true}'
# CLI
npx ai-supply add metagpt-multi-agent-framework
# REST (install → download)
curl -X POST https://ai-supply.store/api/v1/listings/metagpt-multi-agent-framework/install \
-H "Authorization: Bearer $AIM_KEY"
# MCP tool
install_listing({ "slug": "metagpt-multi-agent-framework" })OpenAPI spec →Curated mirror — latest upstream source. See the repository for tagged releases.