multi-agent-framework

Lightweight, production-ready framework for building multi-agent AI systems. Minimal dependencies, maximum flexibility.

Philosophy

Most agent frameworks are either too heavy (LangChain) or too opinionated (AutoGPT). This framework provides the core primitives needed to build robust multi-agent systems without the overhead.

Core Concepts

Agent

from maf import Agent, Tool, Memory

agent = Agent(
    name="researcher",
    model="qwen2.5:14b",        # Local or API
    tools=[Tool.web_search, Tool.code_exec],
    memory=Memory.vector(db="chroma"),
    system_prompt="You are a research specialist...",
)

Task Graph

from maf import TaskGraph, Task

graph = TaskGraph()
recon = graph.add(Task("reconnaissance", agent=scout))
analysis = graph.add(Task("analysis", agent=analyst, depends_on=[recon]))
report = graph.add(Task("report", agent=writer, depends_on=[analysis]))

results = graph.execute(parallel=True)

ACP Protocol

Structured communication between agents:

Agent A ──[TASK_ASSIGN]──► Agent B
Agent B ──[STATUS_UPDATE]──► Agent A
Agent B ──[RESULT]──────────► Agent A
Agent A ──[CRITIQUE]──────── ► Agent B  (optional quality loop)

Supported LLM Backends

• Local: llama.cpp, Ollama (via OpenAI-compatible API)
• Cloud: OpenAI, Anthropic, Mistral, Groq
• Custom: Any OpenAI-compatible endpoint

Installation

pip install git+https://github.com/Dev-next-gen/multi-agent-framework

Directory Structure

multi-agent-framework/
├── maf/
│   ├── agent.py
│   ├── task.py
│   ├── memory.py
│   ├── tools/
│   └── protocols/
├── examples/
│   ├── research_pipeline.py
│   ├── code_review_team.py
│   └── security_audit.py
└── tests/

License

MIT — Léo Camus / NextGen Labs