← All integrations
// integration

A fully local multi-agent team on Ollama

Ollama through the OpenAI-compatible endpoint — no key, no cloud, $0.

Enterprise support
01 setup

A minimal setup.

import { OpenMultiAgent, type AgentConfig } from '@open-multi-agent/core'

const agents: AgentConfig[] = [
  { name: 'researcher', systemPrompt: 'Gather the key facts.' },
  { name: 'writer', systemPrompt: 'Write a tight summary.' },
]

// Fully local — no API key. Requires `ollama serve` running.
const oma = new OpenMultiAgent({
  defaultProvider: 'openai',
  defaultModel: 'llama3.1',
  defaultBaseURL: 'http://localhost:11434/v1',
})
const team = oma.createTeam('brief', { name: 'brief', agents })
const result = await oma.runTeam(team, 'Summarize the latest release notes.')
console.log(result.success)
02 how it fits

How it fits.

Ollama speaks the OpenAI-compatible API, so it’s reached with provider: 'openai' and a local baseURL — no key needed. Tool-calling works with local tool-capable models (Gemma, Llama 3.1, Qwen 3, and others). See the local team walkthrough for a full setup.

// Enterprise

Taking this to production?

open-multi-agent is MIT-licensed and free to run yourself. When you need it delivered, integrated, or supported on a deadline, 元定义科技 (YuanASI) offers commercial delivery and support.

Enterprise support