Run agents as one team, multiply your efficiency in the AI era.

Quick Start | Docs

English | 简体中文 | 繁體中文 | 日本語 | 한국어 | Français | Español

---

One-Minute Getting Started Guide

1. Import a preset team and choose the base agent for each member.
2. Set up workspaces for each member in the team.
3. Message a specific member with @member.

---

🔥 News:

Feature Preview

In v0.4.0, we’ll be introducing a major update:

• Current state: The current free-form collaboration model based on @mentions is too open-ended. Because agent outputs are inherently uncertain, collaboration can be unstable, and the collaboration chain may break at certain points.
• Improvement: We want to introduce a workflow mode to improve collaboration efficiency. This mode uses a centralized communication architecture similar to Claude’s agent teams. Each task becomes a workflow node, and all nodes together form a task-flow DAG. An independent scheduler manages workflow state transitions. For complex tasks, a single planning round may not fully satisfy the requirements, so the workflow should support a second or even third iteration. To make this possible, workflows will include configurable iteration counts and support viewing iteration history.
• We believe this change will significantly improve the practicality of OpenTeams. Stay tuned~

-- 中文 -- 我们将在v0.4.0中进行以下重大更新：

• 现状：当前通过@进行自由协作的方式太开放，因为agent输出具有不确定性，所以协作非常不稳定，协作链会有一定概率断裂；
• 改进：我们希望设计出一种工作流模式来提升协作效率。工作流模式采用中心化的通信架构（类似claude的agent teams），每个任务是一个工作流节点，所有节点组成一个任务流DAG，由一个独立的调度器来管控工作流的状态转移。对于复杂任务，一轮计划可能无法完全满足需求，因此需要进第二轮甚至第三轮迭代，所以这里需要给工作流设置迭代次数，并能查看历史迭代。示意图：
• 这次改动我们相信一定能够让openteams的实用性大幅提升，敬请期待~

Major updates

• 2026.04.14 (v0.3.15)
  • Workspace File Change Viewer
• 2026.04.06 (v0.3.12)
  • Enable dark ui mode
  • fix openteams-cli concurrency issues
• 2026.04.02 (v0.3.10)
  • Implement in-app version update
  • The documentation website is now live.
• 2026.03.24 (v0.3.7):
  • Added the built-in openteams-CLI agent, removing the dependency on a locally installed agent.
  • Fixed a memory leak issue in the executor.

---

Quick Start

Option A: Run with npx

This installation method is recommended for Mac and Linux.

# web
npx openteams-web

Option B: Download Desktop App

Requirements

Starting from v0.3.7, we have built in openteams-cli, so installing the AI Agent is no longer required. You can go to "Settings -> Service Providers" page to configure your API.

⚙️ Refer to the docs for providers' configuration.

---

And You can choose any agent from the list of supported agents.:

Agent	Install
Claude Code	npm i -g @anthropic-ai/claude-code
Gemini CLI	npm i -g @google/gemini-cli
Codex	npm i -g @openai/codex
Qwen Code	npm i -g @qwen-code/qwen-code
OpenCode	npm i -g opencode-ai

📚 More agent installation guides

---

Features

Feature	What you get
Supported agents	supports 10 coding-agent runtimes, including Claude Code, Gemini CLI, Codex, Qwen Code, Amp, Cursor Agent, Copilot, Droid, Kimi Code, and OpenCode. And currently integrating other agents.
Shared group-chat context	Every participant works from the same conversation history instead of juggling copied prompts across separate windows.
Parallel execution	Multiple agents can work on the same task at the same time inside one shared session. Different agents handle the tasks they are best at.
Autonomous collaboration	Agents can @mention each other, hand work off, and coordinate directly inside the chat.
Built-in AI members	Start with 160+ built-in AI members across engineering, marketing, writing, research, and content production.
Built-in AI team presets	Launch with 8 ready-to-use team presets for common workflows.
Team guidelines	Define who leads, who can talk to whom, and how collaboration should happen. Customize your AI team and team guidelines.
Skill library	Equip agents with 1000+ built-in skills, and import your own skills when needed.
Fully local execution	Agents run against your local workspace, and runtime artifacts stay under .openteams/ inside that workspace. There is no need to be concerned about data privacy.

Parallel Agent Execution

Run multiple agents in the same shared context and let them execute in parallel to speed up delivery.

Autonomous Agent Collaboration

openteams lets agents message each other directly without forcing a fixed workflow. If you want more structure, add team guidelines to control communication, appoint a lead agent, or let everyone collaborate freely. The communication pattern is entirely up to your use case.

AI Members

openteams includes 160+ built-in AI members across engineering, marketing, writing, content production, and more. Mix and match them into different teams, customize them, and build role combinations that fit the way you work. We will continue expanding and improving the roster.

AI Teams

openteams ships with 8 built-in team presets for common workflows, so you can get started immediately. We recommend defining team guidelines when you create a team so collaboration stays aligned with how you want the group to operate.

Skill Library

openteams includes 1000+ built-in skills that you can combine and assign to different AI members. You can also import skills you create yourself and apply them directly to your agents. We will keep expanding the skill library with a focus on capabilities that hold up in real production environments.

---

Why we're better

Legend: ✅ Full support | 🟡 Partial support | ❌ No support

Capability	Traditional Single Agent	Multi-window Workflow	Claude Code Agent Team	openteams
Parallelism	❌ No, sequential	🟡 Partial, manual	✅ Yes, Claude subagents	✅ Yes, automatic
Shared context	❌ No	❌ No, copy-paste	🟡 Partial, split subagent contexts	✅ Yes, always in sync
Multi-model collaboration	❌ No	🟡 Partial, manual switching	❌ No, Claude only	✅ Yes, Claude + Gemini + Codex + more
Agent handoff	❌ No	❌ No, you orchestrate it	🟡 Partial, delegated inside Claude	✅ Yes, direct @mentions
Predefined AI member	❌ No	❌ No	❌ No	✅ Yes, 160+ members
Team manager	❌ No	❌ No	❌ No	✅ Yes, Customize team guidelines
Your effort	🔴 High	🔴 Very high	🟠 Medium	🟢 Low

---

Tech Stack

Layer	Technology
Frontend	React + TypeScript + Vite + Tailwind CSS
Backend	Rust
Desktop	Tauri

Local Development

Mac/Linux

# 1. Clone the repository
git clone https://github.com/openteams-lab/openteams.git
cd openteams

# 2. Install dependencies
pnpm i

# 3. Start the development server (runs Rust backend + React frontend)
pnpm run dev

# 4. Build frontend
pnpm --filter frontend build

# 5. Build desktop app
pnpm desktop:build

Windows (PowerShell): Start backend and frontend separately

pnpm run dev cannot run in Windows PowerShell. Use the following commands to run backend and frontend separately.

# 1. Clone the repository
git clone https://github.com/openteams-lab/openteams.git
cd openteams

# 2. Install dependencies
pnpm i

# 3. Generate TypeScript types
pnpm run generate-types

# 4. Run database migrations
pnpm run prepare-db

Terminal A (backend)

$env:FRONTEND_PORT = node scripts/setup-dev-environment.js frontend
$env:BACKEND_PORT = node scripts/setup-dev-environment.js backend
$env:RUST_LOG = "debug"
cargo run --bin server

Terminal B (frontend)

$env:FRONTEND_PORT = <frontend port generated from terminal A>
$env:BACKEND_PORT = <backend port generated from terminal A>
cd frontend
pnpm dev -- --port $env:FRONTEND_PORT --host

Open the frontend page at http://localhost:<FRONTEND_PORT> (example: http://localhost:3001).

Build openteams-cli locally

Use the following commands if you need to compile the local openteams-cli binary instead of using the bundled or published build. the build artifacts will be placed in the binaries directory.

# From the repository root
bun run ./scripts/build-openteams-cli.ts

Release Notes & Roadmap

V0.2

• Multi-agent group chat with shared context
• Parallel agent execution
• Agent @mention and autonomous collaboration
• 10 coding-agent runtimes support (Claude Code, Gemini CLI, Codex, Qwen Code, Amp, Cursor Agent, Copilot, Droid, Kimi Code, OpenCode)
• Desktop apps (Windows, macOS, Linux)
• Web app via npx
• Multi-language support (EN, ZH, JA, KO, FR, ES)

V0.3

• The frontend interface has been completely overhauled.
• 160+ built-in AI members
• 8 built-in AI team presets
• Team rules configuration
• 1000+ built-in skills
• Fully local execution with workspace isolation
• Redefine the input protocol

Roadmap

• Backend for Code Agent optimized for OpenTeams use cases —— v0.3.7
• Develop multiple frontend color schemes —— v0.3.11
• Establish a high-efficiency team collaboration framework
• More agent integrations (Kilo Code, OpenClaw, etc.)
• Add more powerful, ready-to-use AI teams
• Added more powerful skills
• Offer a highly optimized, customized version

Contributing

We welcome contributions! Check what's needed in Issues or start a Discussion.

1. Fork -> feature branch -> PR
2. Please open an issue before large changes
3. Please follow our Code of Conduct

Code Formatting

Before submitting a PR, please ensure your code is properly formatted:

# Format both frontend and backend
pnpm run format

# Check formatting without making changes
pnpm run format:check

# Format frontend only
pnpm run frontend:format

# Format backend only (Rust)
pnpm run backend:format

Note: CI will fail if code formatting is not correct. Always run pnpm run format:check before pushing.

Community

Bug Reports	GitHub Issues
Discussions	GitHub Discussions
Community Chat	Discord

Acknowledgements

Built on top of Vibe Kanban - thanks to their team for the excellent open source foundation.

Thanks as well to ComposioHQ/awesome-claude-skills for helping shape the built-in skill ecosystem, and to msitarzewski/agency-agents for inspiration around agent role design and team composition.