Kickstart Multi-Agent Collaboration with OxyGent: 20‑Line Setup Guide

OxyGent Multi-Agent Collaboration Framework

Open‑source repository: GitHub . Official site: oxygent.jd.com . PyPI package: oxygent .

Quick Start in 20 Lines of Code

Installation

Python environment (3.10+)

conda create -n oxy_env python==3.10

conda activate oxy_env

OxyGent package pip install oxygent Node.js (if using MCP tools) – download from nodejs.org

Hello World

Create a .env file and set the large‑model environment variables:

DEFAULT_LLM_API_KEY    = "<large‑model‑key>"

DEFAULT_LLM_BASE_URL    = "<large‑model‑url>"

DEFAULT_LLM_MODEL_NAME = "<large‑model‑name>"

Run the demo (see image).

RAG

MoA

Enable Agents to Call Tools Autonomously

SSE MCP Tool

FunctionHub Tool

Using any registration method (FunctionHub, LocalMCP, SSEMCP) yields the same runtime effect.

External MCP Tool

Auto‑recall Top‑K Tools

Enter the node visualization page:

Block‑Based Multi‑Agent Construction

Multi‑Agent

Hierarchical Agents

Workflow Integration

Reflexion Mechanism

Why?

Fast Agent Deployment

Data Persistence

The framework provides a complete data storage mechanism for later SFT or RL training.

Limit Node Concurrency

Multi‑Environment Deployment

Distributed

Advanced Usage

Multimodal

Weight‑based Memory filtering

Retrieve more tools

Custom LLM output parser

Custom SSE interface

Result post‑processing or formatting

Agents calling multiple tools simultaneously

Restart tasks from intermediate nodes

Plan‑and‑Solve paradigm

…