Open Multi Agent

v1.0.0Coding Agentsstable

From a goal to a task DAG, automatically. TypeScript-native multi-agent orchestration with MCP and live tracing. Three runtime dependencies.

agent-frameworkai-agentsanthropicclaudedeepseek
Share:
6,218
Stars
0
Downloads
0
Weekly
0/5

What is Open Multi Agent?

Open Multi Agent is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to from a goal to a task dag, automatically. typescript-native multi-agent orchestration with mcp and live tracing. three runtime dependencies.

From a goal to a task DAG, automatically. TypeScript-native multi-agent orchestration with MCP and live tracing. Three runtime dependencies.

This server falls under the Coding Agents category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • From a goal to a task DAG, automatically. TypeScript-native

Use Cases

Multi-agent orchestration
Task DAG generation
Live tracing
LicenseMIT
Languagetypescript
Versionv1.0.0
UpdatedMay 22, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx open-multi-agent

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Open Multi Agent

Open Multi-Agent is a TypeScript-native orchestration framework that converts a high-level goal into a directed acyclic graph (DAG) of tasks and executes them with a coordinated team of AI agents. It supports twelve built-in LLM providers — including Anthropic, OpenAI, Gemini, DeepSeek, and local models via Ollama — while exposing filesystem tools, custom tool definitions, and the ability to consume any stdio-based MCP server. Developers use it to build autonomous pipelines where sub-agents collaborate, hand off work, and trace execution in a live HTML dashboard.

Prerequisites

  • Node.js 18 or later installed
  • At least one LLM provider API key (e.g., ANTHROPIC_API_KEY, OPENAI_API_KEY, or GEMINI_API_KEY)
  • npm or npx available on your PATH
  • An MCP client such as Claude Desktop or Claude Code if consuming via MCP
1

Install the core package

Add the @open-multi-agent/core package to your TypeScript or JavaScript project. Node.js 18+ is required.

npm install @open-multi-agent/core
2

Set your provider API key

Export the API key for your chosen LLM provider. Open Multi-Agent supports Anthropic, OpenAI, Gemini, Azure OpenAI, DeepSeek, Groq, Mistral, and many others.

export ANTHROPIC_API_KEY=sk-ant-your-key-here
# or
export OPENAI_API_KEY=sk-your-key-here
# or
export GEMINI_API_KEY=your-gemini-key
3

Define an agent configuration

Create an AgentConfig object specifying the model to use and which built-in tools to grant. Tools are default-deny and must be explicitly listed.

import { AgentConfig } from '@open-multi-agent/core';

const agent: AgentConfig = {
  name: 'developer',
  model: 'claude-sonnet-4-6',
  tools: ['bash', 'file_read', 'file_write', 'grep', 'glob'],
};
4

Run a single agent or a full team

Use runAgent for a single focused task, or runTeam to automatically decompose a goal into a DAG of parallel and sequential sub-tasks.

import { Orchestrator } from '@open-multi-agent/core';

const orchestrator = new Orchestrator();

// Single agent
await orchestrator.runAgent(agent, 'Refactor auth.ts to use async/await');

// Auto-DAG team run
await orchestrator.runTeam(team, 'Build a REST API with tests and documentation');
5

Connect an MCP server as a tool source

Open Multi-Agent can consume any stdio-based MCP server, making its tools available to agents in your pipeline.

import { connectMCPTools } from '@open-multi-agent/core';

const mcpTools = await connectMCPTools({
  command: 'npx',
  args: ['-y', '@wonderwhy-er/desktop-commander@latest'],
});

const agent: AgentConfig = {
  name: 'file-agent',
  model: 'claude-sonnet-4-6',
  tools: [...mcpTools],
};
6

Configure resource limits and observability

Set token budgets, timeouts, retry counts, and context strategies to make long-running pipelines reliable. Enable onProgress and onTrace callbacks for live observability.

const orchestrator = new Orchestrator({
  maxTokenBudget: 100000,
  timeoutMs: 300000,
  maxRetries: 3,
  retryDelayMs: 2000,
  contextStrategy: 'summarize',
  onProgress: (event) => console.log(event),
});

Open Multi Agent Examples

Client configuration

Open Multi-Agent is a framework library, not an MCP server itself. Use it in a Node.js script that you run directly. Here is a minimal package.json and entry point.

// package.json
{
  "name": "my-agent-pipeline",
  "type": "module",
  "scripts": { "start": "node agent.js" },
  "dependencies": { "@open-multi-agent/core": "latest" }
}

// agent.js
import { Orchestrator } from '@open-multi-agent/core';
const orch = new Orchestrator();
await orch.runAgent(
  { name: 'coder', model: 'claude-sonnet-4-6', tools: ['file_read', 'file_write', 'bash'] },
  process.argv[2] || 'List all TypeScript files in src/'
);

Prompts to try

These goal strings work well as the prompt argument to runAgent or runTeam.

- "Analyse the performance bottlenecks in src/ and suggest optimisations"
- "Write unit tests for all exported functions in lib/utils.ts"
- "Create a three-agent pipeline: researcher, writer, and reviewer to produce a technical blog post about MCP"
- "Decompose building a Node.js CLI tool into parallel tasks and execute them"

Troubleshooting Open Multi Agent

Error: model not found or API key invalid

Ensure the correct environment variable is exported for your provider (e.g., ANTHROPIC_API_KEY for claude-* models). Run `echo $ANTHROPIC_API_KEY` to confirm the variable is set in the current shell.

Agents exceed token budget and stop early

Increase maxTokenBudget in the Orchestrator config, or set contextStrategy to 'summarize' or 'compact' to compress conversation history between turns.

MCP tools not available to agents after connectMCPTools

Confirm the MCP server process starts successfully by running its command manually (e.g., npx -y @wonderwhy-er/desktop-commander@latest). Spread the returned tools array into the agent's tools field rather than assigning it directly.

Frequently Asked Questions about Open Multi Agent

What is Open Multi Agent?

Open Multi Agent is a Model Context Protocol (MCP) server that from a goal to a task dag, automatically. typescript-native multi-agent orchestration with mcp and live tracing. three runtime dependencies. It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Open Multi Agent?

Follow the installation instructions on the Open Multi Agent GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with Open Multi Agent?

Open Multi Agent works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Open Multi Agent free to use?

Yes, Open Multi Agent is open source and available under the MIT license. You can use it freely in both personal and commercial projects.

Browse More Coding Agents MCP Servers

Explore all coding agents servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "open-multi-agent": { "command": "npx", "args": ["-y", "open-multi-agent"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Open Multi Agent?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides