MCPInstall

open-multi-agent-orchestration

by reason-machinesAI & Agents

TypeScript-native multi-agent orchestration framework that decomposes goals into task DAGs automatically with MCP and live tracing

Installs
197
Last 24h
—
Trending rank
—
Security
Not audited

Install

Works with every agent the skills CLI supports.

npx skills add https://github.com/reason-machines/ai-agent-skills --skill open-multi-agent-orchestration
  1. 01

    Install

    Run the command in your project. The CLI asks which agents to install for.

  2. 02

    Describe the task

    Ask your agent for something this skill covers. It loads open-multi-agent-orchestration automatically when relevant.

  3. 03

    Review the result

    The agent follows the SKILL.md steps, scripts and references bundled with the skill.

SKILL.md

The instructions your agent reads when the skill is triggered.

Open Multi-Agent Orchestration

Skill by ara.so — AI Agent Skills collection.

Open Multi-Agent is a TypeScript-native multi-agent orchestration framework that automatically decomposes goals into task DAGs, parallelizes independent tasks, and synthesizes results. It supports 10+ LLM providers, built-in tools, MCP server integration, and has only three runtime dependencies.

Installation

npm install @open-multi-agent/core

Requirements: Node.js >= 18

Core Concepts

Three Execution Modes

  1. Single Agent - One agent, one prompt
  2. Auto-orchestrated Team - Coordinator decomposes goal into tasks automatically
  3. Explicit Pipeline - You define the task graph and assignments

Basic Single Agent

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

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6',
})

const result = await orchestrator.runAgent({
  name: 'coder',
  systemPrompt: 'You are an expert TypeScript developer.',
  tools: ['bash', 'file_write', 'file_read'],
}, 'Create a simple Express server in /tmp/api')

console.log(result.success) // true
console.log(result.content) // agent's final response
console.log(result.totalTokenUsage) // { input_tokens: 1234, output_tokens: 567 }

Auto-Orchestrated Team (Recommended)

The coordinator agent decomposes your goal into a task DAG and executes it:

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

const agents: AgentConfig[] = [
  {
    name: 'architect',
    model: 'claude-sonnet-4-6',
    systemPrompt: 'Design clean API contracts and data models.',
    tools: ['file_write'],
  },
  {
    name: 'developer',
    model: 'claude-sonnet-4-6',
    systemPrompt: 'Implement runnable TypeScript code.',
    tools: ['bash', 'file_read', 'file_write', 'file_edit'],
  },
  {
    name: 'reviewer',
    model: 'claude-sonnet-4-6',
    systemPrompt: 'Review code for correctness and security.',
    tools: ['file_read', 'grep'],
  },
]

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6',
  onProgress: (event) => {
    console.log(event.type, event.task ?? event.agent ?? '')
  },
})

const team = orchestrator.createTeam('api-team', {
  name: 'api-team',
  agents,
  sharedMemory: true, // agents can share context
})

const result = await orchestrator.runTeam(
  team,
  'Create a REST API for a todo list in /tmp/todo-api/'
)

console.log(result.success)
console.log(result.content) // synthesized final result
console.log(result.totalTokenUsage.output_tokens)

Explicit Task Pipeline

When you know the exact workflow:

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

const tasks: TaskConfig[] = [
  {
    id: 'design',
    description: 'Design the API schema',
    assignedTo: 'architect',
    dependencies: [],
  },
  {
    id: 'implement',
    description: 'Implement the endpoints',
    assignedTo: 'developer',
    dependencies: ['design'],
  },
  {
    id: 'test',
    description: 'Write integration tests',
    assignedTo: 'developer',
    dependencies: ['implement'],
  },
  {
    id: 'review',
    description: 'Security and code review',
    assignedTo: 'reviewer',
    dependencies: ['implement', 'test'],
  },
]

const result = await orchestrator.runTasks(team, tasks)

Provider Configuration

Environment Variables

# Anthropic
export ANTHROPIC_API_KEY=sk-ant-...

# OpenAI
export OPENAI_API_KEY=sk-...

# Google Gemini
export GEMINI_API_KEY=...

# DeepSeek
export DEEPSEEK_API_KEY=sk-...

# Azure OpenAI
export AZURE_OPENAI_API_KEY=...
export AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com
export AZURE_OPENAI_DEPLOYMENT_NAME=gpt-4
export AZURE_OPENAI_API_VERSION=2024-02-15-preview

# Ollama (local)
# No API key needed, runs on localhost:11434 by default

Using Multiple Providers in One Team

const agents: AgentConfig[] = [
  {
    name: 'planner',
    model: 'claude-sonnet-4-6', // Anthropic
    systemPrompt: 'Create detailed plans.',
  },
  {
    name: 'coder',
    model: 'gpt-4o', // OpenAI
    systemPrompt: 'Write production-grade code.',
    tools: ['bash', 'file_write'],
  },
  {
    name: 'local-reviewer',
    model: 'ollama:qwen2.5-coder:32b', // Local Ollama
    systemPrompt: 'Review code for bugs.',
    tools: ['file_read', 'grep'],
  },
]

Ollama (Local Models)

const orchestrator = new OpenMultiAgent({
  defaultModel: 'ollama:qwen2.5-coder:7b',
})

const result = await orchestrator.runAgent({
  name: 'local-coder',
  model: 'ollama:deepseek-coder-v2:16b',
  systemPrompt: 'You write Python code.',
  tools: ['bash', 'file_write'],
}, 'Create a FastAPI hello world in /tmp/api.py')

Tools

Built-in Tools

Available out of the box:

  • bash - Execute shell commands
  • file_read - Read file contents
  • file_write - Write files
  • file_edit - Edit files using search/replace
  • grep - Search file contents
  • glob - List files matching patterns
const agent: AgentConfig = {
  name: 'dev',
  systemPrompt: 'You are a developer.',
  tools: ['bash', 'file_read', 'file_write', 'file_edit', 'grep'],
}

Custom Tools with Zod

import { defineTool } from '@open-multi-agent/core'
import { z } from 'zod'

const weatherTool = defineTool({
  name: 'get_weather',
  description: 'Get current weather for a city',
  parameters: z.object({
    city: z.string().describe('City name'),
    units: z.enum(['celsius', 'fahrenheit']).default('celsius'),
  }),
  execute: async ({ city, units }) => {
    // Your implementation
    const temp = units === 'celsius' ? 22 : 72
    return `Weather in ${city}: ${temp}°${units === 'celsius' ? 'C' : 'F'}`
  },
})

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6',
  customTools: [weatherTool],
})

const result = await orchestrator.runAgent({
  name: 'assistant',
  tools: ['get_weather'],
}, 'What is the weather in London?')

MCP Server Integration

Connect Model Context Protocol servers:

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

const mcpTools = await connectMCPTools({
  command: 'npx',
  args: ['-y', '@modelcontextprotocol/server-github'],
  env: {
    GITHUB_PERSONAL_ACCESS_TOKEN: process.env.GITHUB_TOKEN,
  },
})

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6',
  customTools: mcpTools,
})

const result = await orchestrator.runAgent({
  name: 'github-agent',
  tools: ['create_or_update_file', 'search_repositories'], // MCP tools
}, 'Search for TypeScript agent frameworks and create a comparison in repo.md')

Agent Delegation Tool

Allow agents to delegate to other agents:

const team = orchestrator.createTeam('dev-team', {
  name: 'dev-team',
  agents: [
    {
      name: 'lead',
      systemPrompt: 'You coordinate work.',
      tools: ['delegate_to_agent'],
    },
    {
      name: 'specialist',
      systemPrompt: 'You implement features.',
      tools: ['bash', 'file_write'],
    },
  ],
})

// The 'lead' agent can now call delegate_to_agent to hand off tasks

Structured Output

Get Zod-validated responses:

import { z } from 'zod'

const resultSchema = z.object({
  files: z.array(z.object({
    path: z.string(),
    purpose: z.string(),
  })),
  commands: z.array(z.string()),
  summary: z.string(),
})

const result = await orchestrator.runAgent({
  name: 'architect',
  systemPrompt: 'You design project structures.',
}, 'Design a TypeScript library structure', {
  resultSchema,
})

// result.parsedContent is now typed and validated
console.log(result.parsedContent.files) // TypeScript knows the shape
console.log(result.parsedContent.commands)

Shared Memory

In-Memory (Default)

const team = orchestrator.createTeam('team', {
  name: 'team',
  agents: [...],
  sharedMemory: true, // default in-memory store
})

Custom Memory Store (Redis)

import { MemoryStore } from '@open-multi-agent/core'
import Redis from 'ioredis'

class RedisMemoryStore implements MemoryStore {
  private redis: Redis

  constructor() {
    this.redis = new Redis(process.env.REDIS_URL)
  }

  async get(key: string): Promise<string | null> {
    return this.redis.get(key)
  }

  async set(key: string, value: string): Promise<void> {
    await this.redis.set(key, value)
  }

  async delete(key: string): Promise<void> {
    await this.redis.del(key)
  }

  async clear(): Promise<void> {
    await this.redis.flushdb()
  }
}

const team = orchestrator.createTeam('team', {
  name: 'team',
  agents: [...],
  sharedMemory: true,
  memoryStore: new RedisMemoryStore(),
})

Observability

Progress Events

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6',
  onProgress: (event) => {
    switch (event.type) {
      case 'agent_start':
        console.log(`Starting agent: ${event.agent}`)
        break
      case 'task_start':
        console.log(`Task started: ${event.task}`)
        break
      case 'task_complete':
        console.log(`Task complete: ${event.task}`)
        break
      case 'agent_complete':
        console.log(`Agent finished: ${event.agent}`)
        break
    }
  },
})

Trace Observability

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6',
  onTrace: (span) => {
    console.log('Span:', span.type, span.name)
    console.log('Duration:', span.endTime - span.startTime, 'ms')
    if (span.metadata?.tokens) {
      console.log('Tokens:', span.metadata.tokens)
    }
  },
})

Post-Run Dashboard

Generate HTML report of executed task DAG:

const result = await orchestrator.runTeam(team, goal)

// result.trace contains all execution data
// Render to HTML (implementation in docs/observability.md)

Plan-Only Mode

Preview the task DAG without executing:

const plan = await orchestrator.runTeam(team, goal, { planOnly: true })

console.log('Tasks:', plan.tasks.map(t => ({
  id: t.id,
  description: t.description,
  assignedTo: t.assignedTo,
  dependencies: t.dependencies,
})))

Fan-Out / MapReduce Pattern

Parallel execution without dependencies:

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

const pool = new AgentPool({
  defaultModel: 'claude-sonnet-4-6',
  agents: [
    { name: 'analyzer-1', systemPrompt: 'Analyze data source 1.' },
    { name: 'analyzer-2', systemPrompt: 'Analyze data source 2.' },
    { name: 'analyzer-3', systemPrompt: 'Analyze data source 3.' },
  ],
})

const results = await pool.runParallel([
  { agent: 'analyzer-1', prompt: 'Analyze feed A' },
  { agent: 'analyzer-2', prompt: 'Analyze feed B' },
  { agent: 'analyzer-3', prompt: 'Analyze feed C' },
])

// Aggregate results
const aggregator = new OpenMultiAgent({ defaultModel: 'claude-sonnet-4-6' })
const summary = await aggregator.runAgent({
  name: 'aggregator',
  systemPrompt: 'Merge analysis results.',
}, `Merge these findings:\n${results.map(r => r.content).join('\n\n')}`)

Production Checklist

Context Management

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6',
  maxContextTokens: 100000, // prevent context overflow
  contextStrategy: 'truncate', // or 'summarize'
})

Task Retry with Backoff

const tasks: TaskConfig[] = [
  {
    id: 'api-call',
    description: 'Fetch external data',
    assignedTo: 'fetcher',
    dependencies: [],
    retry: {
      maxAttempts: 3,
      backoffMs: 1000, // exponential backoff
    },
  },
]

Loop Detection

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6',
  maxIterations: 20, // prevent infinite loops
})

Tool Output Truncation

import { defineTool } from '@open-multi-agent/core'
import { z } from 'zod'

const largeTool = defineTool({
  name: 'fetch_logs',
  description: 'Fetch system logs',
  parameters: z.object({ lines: z.number() }),
  execute: async ({ lines }) => {
    const logs = getLogs(lines) // potentially huge
    return logs.slice(0, 10000) // truncate to 10k chars
  },
})

Common Patterns

Contract Review Workflow

Four-task DAG with parallel branches:

const tasks: TaskConfig[] = [
  {
    id: 'extract',
    description: 'Extract key terms from contract',
    assignedTo: 'extractor',
    dependencies: [],
  },
  {
    id: 'risk-check',
    description: 'Identify legal risks',
    assignedTo: 'legal-expert',
    dependencies: ['extract'],
  },
  {
    id: 'compliance-check',
    description: 'Check regulatory compliance',
    assignedTo: 'compliance-expert',
    dependencies: ['extract'],
  },
  {
    id: 'report',
    description: 'Generate final report',
    assignedTo: 'reporter',
    dependencies: ['risk-check', 'compliance-check'],
  },
]

const result = await orchestrator.runTasks(team, tasks)

Translation with Back-Translation

const team = orchestrator.createTeam('translation', {
  name: 'translation',
  agents: [
    {
      name: 'translator',
      model: 'claude-sonnet-4-6',
      systemPrompt: 'Translate English to target language.',
    },
    {
      name: 'back-translator',
      model: 'gpt-4o',
      systemPrompt: 'Translate back to English.',
    },
    {
      name: 'validator',
      model: 'claude-sonnet-4-6',
      systemPrompt: 'Compare original and back-translation, flag drift.',
    },
  ],
  sharedMemory: true,
})

const tasks: TaskConfig[] = [
  { id: 'translate', description: 'Translate to Spanish', assignedTo: 'translator', dependencies: [] },
  { id: 'back-translate', description: 'Translate back to English', assignedTo: 'back-translator', dependencies: ['translate'] },
  { id: 'validate', description: 'Compare and flag drift', assignedTo: 'validator', dependencies: ['translate', 'back-translate'] },
]

const result = await orchestrator.runTasks(team, tasks)

Troubleshooting

"Model not found" Error

Make sure the correct API key is set:

# Check which provider your model uses
export ANTHROPIC_API_KEY=sk-ant-...  # for claude-*
export OPENAI_API_KEY=sk-...         # for gpt-*
export GEMINI_API_KEY=...            # for gemini-*

Ollama Connection Refused

# Start Ollama daemon
ollama serve

# Pull the model
ollama pull qwen2.5-coder:7b

# Verify it's running
curl http://localhost:11434/api/tags

Task Hangs / Times Out

Enable progress logging to see where it's stuck:

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6',
  onProgress: (event) => console.log(event), // debug all events
  timeout: 300000, // 5 minutes per task
})

"Too many tokens" Error

Reduce context or use a larger context window model:

const orchestrator = new OpenMultiAgent({
  defaultModel: 'claude-sonnet-4-6', // 200k context
  maxContextTokens: 100000,
  contextStrategy: 'truncate', // keep recent messages only
})

MCP Server Won't Connect

Check stderr from the MCP process:

const mcpTools = await connectMCPTools({
  command: 'npx',
  args: ['-y', '@modelcontextprotocol/server-github'],
  env: { GITHUB_PERSONAL_ACCESS_TOKEN: process.env.GITHUB_TOKEN },
  onStderr: (data) => console.error('MCP stderr:', data),
})

Agents Not Sharing Context

Ensure sharedMemory: true is set:

const team = orchestrator.createTeam('team', {
  name: 'team',
  agents: [...],
  sharedMemory: true, // required for context sharing
})

CLI Usage

For shell and CI workflows:

# Install globally
npm install -g @open-multi-agent/core

# Run a team
oma run --team api-team --goal "Create REST API" --config team.json

# Plan only
oma plan --team api-team --goal "Create REST API" --config team.json

# Output JSON for parsing
oma run --team api-team --goal "..." --json > result.json

See docs/cli.md for full CLI reference.

Resources

FAQ

What does the open-multi-agent-orchestration skill do?

TypeScript-native multi-agent orchestration framework that decomposes goals into task DAGs automatically with MCP and live tracing

How do I install open-multi-agent-orchestration?

Run npx skills add https://github.com/reason-machines/ai-agent-skills --skill open-multi-agent-orchestration in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/open-multi-agent-orchestration/ (personal) or .claude/skills/open-multi-agent-orchestration/ (project).

Which AI agents can use open-multi-agent-orchestration?

Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.

Is open-multi-agent-orchestration safe to install?

No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.

Who maintains open-multi-agent-orchestration?

It is published by reason-machines in reason-machines/ai-agent-skills. It has 197 recorded installs.

More from reason-machines/ai-agent-skills

View all →

agency-agents-zh-skill

reason-machines/ai-agent-skills

Open the skill to read its SKILL.md instructions.

AI & Agents322 installs

agent-reach-internet-access

reason-machines/ai-agent-skills

Open the skill to read its SKILL.md instructions.

AI & Agents228 installs

agent-sprite-forge-2d-game-assets

reason-machines/ai-agent-skills

Open the skill to read its SKILL.md instructions.

Media & Creative216 installs

agentmemory-persistent-memory

reason-machines/ai-agent-skills

Add persistent memory to AI coding agents using agentmemory - remembers context, preferences, and decisions across sessions

AI & Agents216 installs

awesome-agent-skills

reason-machines/ai-agent-skills

Open the skill to read its SKILL.md instructions.

AI & Agents214 installs

oh-my-openagent-orchestration

reason-machines/ai-agent-skills

Open the skill to read its SKILL.md instructions.

AI & Agents211 installs

Popular ai & agents skills

View category →

google-agents-cli-publish

google/agents-cli

This skill should be used when the user wants to "publish an agent", "publish my ADK agent", "register an agent with Gemini", "publish to Gemini" (formerly Gemini Enterprise), or needs guidance on the agents-cli publish gemini command. Also use when the user wants to "manage agents in Agent Registry", "list/update/delete registered agents", or "register an MCP server". Covers ADK vs A2A registration modes, programmatic and interactive usage, flag reference, auto-detection from deployment metadata, Agent Registry fleet management, and troubleshooting. Part of the agents-cli skills suite. Do NOT use for deployment (use google-agents-cli-deploy).

AI & Agents513.9K installs

google-agents-cli-eval

google/agents-cli

This skill should be used when the user wants to "run an evaluation", "evaluate my agent", "evaluate my ADK agent", "write an eval dataset", "analyze eval failures", "compare eval results", "optimize agent", or needs guidance on the Gemini platform eval methodology and the Quality Flywheel. Covers eval metrics, dataset schema, LLM-as-judge scoring, and common failure causes. Applies to any agents-cli project, whatever framework the agent is written in. Do NOT use for agent API code patterns (ADK: use google-agents-cli-adk-code), deployment (use google-agents-cli-deploy), or project scaffolding (use google-agents-cli-scaffold).

AI & Agents513.5K installs

skill-creator

anthropics/skills

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

OfficialAI & Agents416.4K installs

codex-pet

prime-skills/runcomfy-agent-skills

Codex Pet generator on RunComfy. Build a Codex-compatible Codex Pet spritesheet.webp + pet.json from a single reference image, drop it into ${CODEXHOME:-$HOME/.codex}/pets/ / and Codex picks it up as a custom Codex Pet next to the 8 built-ins. This skill produces the exact Codex Pet atlas Codex expects (1536x1872 PNG/WebP, 8 cols x 9 rows, 192x208 cells, 9 animation states — idle, running-right, running-left, waving, jumping, failed, waiting, running, review). Calls OpenAI GPT Image 2 edit ONCE via the local RunComfy CLI as runcomfy run openai/gpt-image-2/edit to produce a canonical Codex Pet pose, then assembles all 9 animation rows programmatically with ImageMagick micro-transforms — no Codex Pro, no $imagegen, no OPENAIAPIKEY required, only RUNCOMFYTOKEN. Triggers on "codex pet", "create codex pet", "make codex pet", "hatch codex pet", "/hatch image", "desktop pet codex", "codex pets", "spritesheet.webp", or any explicit ask to build a custom pet for OpenAI Codex.

AI & Agents402.1K installs

full-output-enforcement

leonxlnx/taste-skill

Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.

AI & Agents401K installs

executing-plans

obra/superpowers

Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available

AI & Agents245.7K installs