Plan, Build, and Ship — with a team of AI agents instead of one
openteams is an open-source, local-first AI desktop app that helps indie developers plan, build, and ship software faster with an AI team they can control, supporting 13 coding agents including Claude Code, Codex, Gemini CLI, and the bundled openteams-cli.
What is openteams?
You already run Claude Code, Codex, Gemini CLI, or another coding agent. Each one works well on its own. Then you open a second terminal, and a third. You repeat the same context, carry results from one window to another, and keep track of who is changing what. Before long, you are managing the agents instead of the work: changes are scattered across sessions, project priorities live elsewhere, and token usage is disconnected from what actually shipped.
openteams is what is missing around those agents: one shared room where they can talk and hand off work, a plan you can see and control, and a lightweight local record that connects project work to agent output without giving agents control of the roadmap.
| openteams is | openteams is not |
|---|---|
| a local-first workspace for the coding agents you already use | another model or a replacement for Claude Code, Codex, or Gemini CLI |
| a shared session where agents can talk, hand off work, and keep the same context | a pile of separate chat windows that you have to coordinate yourself |
| a developer-owned issue list linked to agent sessions | a full project-management suite or a roadmap that agents rewrite themselves |
| a workflow you can inspect, review, interrupt, and retry step by step | one large prompt that disappears into a black box until it finishes |
| isolated worktrees you can review, merge, or discard separately | multiple agents changing the same workspace and interfering with each other |
| build statistics that show what agents delivered, the resources they used, and what it cost | a token counter with no record of what was built |
Concretely, installing it gives you: chat sessions for lightweight collaboration and planned execution, ready-to-use team workflow templates, developer-controlled Issues that link work to sessions, independent workspaces for isolating parallel tasks, and complete build statistics.
without openteams with openteams
Claude ─ terminal A ─┐ Claude ─┐
Codex ── terminal B ─┼─ you relay Codex ──┼─ shared session
Gemini ─ terminal C ─┘ Gemini ─┘
plan: somewhere else issues ── sessions ── build results
Why openteams
AI coding agents can produce plenty of code. The harder part is keeping their work organized: sharing context, seeing what is happening, stopping parallel tasks from overwriting each other, deciding what comes next, and knowing what the work cost.
openteams keeps agents and their conversations in one session. For larger tasks, Workflow mode shows the steps and dependencies before and during execution, so you can review or retry one step without starting everything again. If several sessions are working at once, each can use its own Git worktree, keeping unfinished changes separate until you choose to merge or discard them.
The project direction stays with the developer. Issues hold the work you have chosen and link it to the sessions where agents carry it out; agents do the work, but they do not rewrite the plan. Build statistics then show what was delivered alongside the tokens and cost used to deliver it.
openteams is not trying to give you more agents. It is trying to make sure you always know what is being built, where the changes are, what comes next, and what it cost.
Quick Start
Install
Desktop App (Recommended)
Download the latest release for your platform from GitHub Releases.
macOS: The current macOS release is not signed or notarized by Apple. Browsers add a quarantine attribute to apps downloaded from the internet, so Gatekeeper may report that openteams is “damaged” even when the download is intact. After dragging openteams.app to /Applications, and only if you trust that it came from the official openteams GitHub Release, run:
xattr -dr com.apple.quarantine /Applications/openteams.app
This removes the quarantine attribute only from openteams; it does not disable Gatekeeper globally.
npx
npx openteams-web
Configure Providers
openteams includes a built-in openteams CLI agent. Configure your model providers in the app under Settings → Provider Config → Add Provider.
You can also connect any of these supported coding agents:
| Agent | Example install |
|---|---|
| Claude Code | npm i -g @anthropic-ai/claude-code |
| Codex | npm i -g @openai/codex |
| Gemini CLI | npm i -g @google/gemini-cli |
| Qwen Code | npm i -g @qwen-code/qwen-code |
| OpenCode | npm i -g opencode-ai |
| Amp | npm i -g @sourcegraph/amp |
| GitHub Copilot | npm i -g @github/copilot |
| Cursor | curl https://cursor.com/install -fsS | bash |
| Droid | curl -fsSL https://app.factory.ai/cli | sh |
| Kimi Code | curl -LsSf https://code.kimi.com/install.sh | bash |
| Qoder CLI | curl -fsSL https://qoder.com/install | bash |
| Pi | No install needed (runs via npx) |
| DeepSeek Harness (ACP preview) | Build the official source checkout using the in-app pinned setup commands; OpenTeams launches its ACP server directly with Node (no npx) |
| OpenTeams CLI | Bundled with openteams |
Major updates
- 2026.08.04 (v1.0.9)
- Support Pi agent (new ACP-based runtime with approval, launcher, and MCP extensions)
- Workflow runtime hardening: prompt safety, budget caps, reviewer-loop invariants, and per-prompt budget allocation
- Redesign chat agent activity log for readability; render activity log prose with shared markdown renderer
- Support Qoder CLI; format workflow detail and log UI
- Agent install notify and OAuth status checking
- 2026.07.27 (v1.0.5)
- Support ACP approval and agent permission config
- Add workflow node skip functionality
- Universe app executor v1.0
- Improve permission UI and inbox notification style
- Transform Kimi to ACP protocol; manage
~/.openteams/opencodenumber and size
- 2026.07.20 (v1.0.0)
- First stable release: custom AI members, agent auto mode, merge conflict window
- macOS DMG installer styling; command help dialog for new users
- Support visiting links in agent messages and Enter-to-confirm dialogs
- 2026.05.20 (v0.4.4)
- Workflow mode beta version
Roadmap
openteams is under active development. Here is where we are heading:
- Expert AI workers — Launch more AI workers with deep domain knowledge that can solve specialized problems.
- High-output AI teams — Compose efficient expert AI workers into teams that can customize production workflows for specific business needs and turn requirements into deliverables end to end.
- Integrate more agents — Add support for more commonly used agents, such as Kilo Code, hermes-agent, openclaw, and others.
Vision: Transform token consumption into real productivity.
Have a feature request or want to help shape the direction? Open a discussion.
Community
- GitHub Issues: bug reports and feature requests
- GitHub Discussions: product ideas and questions
- Discord: community chat
- Linux.do: friendly link; thanks for providing community discussion support
- Community groups:
Core Features
| Feature | What it means |
|---|---|
| AI employees and AI teams | Turn tokens into real productivity. Each AI employee or team carries domain-specific expertise that elevates general-purpose models into specialists — ready to ship work, not just generate text. |
| Multi-agent workspace | Bring multiple AI agents into one shared session instead of juggling separate windows. |
| Shared context | Agents work from the same conversation and project context. |
| Free Chat | Use @ for direct, lightweight agent collaboration. |
| Workflow mode | Convert complex tasks into structured steps, dependencies, reviews, retries, and acceptance. |
| Visible execution | See what each agent is doing and where the work is blocked. |
| Review and retry | Review a step, retry the right task, and avoid restarting the whole project. |
| Issue management | Record and prioritize developer-controlled work items, sync issues from GitHub, and create or link sessions for execution. |
| Isolated workspaces | Run independent session tasks in separate Git worktrees, then review, merge, or discard each result without interfering with other work. |
| Build statistics | Compare bugs fixed and features delivered with token usage and cost breakdowns across sessions and models. |
| Artifacts and traces | Keep logs, diffs, transcripts, and generated artifacts attached to the work. |
| Local workspace execution | Agents work against your configured workspace, with runtime records kept under .openteams/. |
Who It Is For
openteams is for:
- developers using multiple coding agents who are tired of juggling them
- technical leads who need agent runs to be reviewable and reproducible
It is not just a place to collect more agents. It is a way to turn agents into a working team.
Tech Stack
| Layer | Technology |
|---|---|
| Frontend | React, TypeScript, Vite, Tailwind CSS |
| Backend | Rust |
| Desktop | Tauri |
| Database | SQLx-managed relational schema |
| Workflow UI | React Flow |
Local Development
Prerequisites
- Rust >= 1.75
- Node.js >= 18
- pnpm >= 8
macOS, Linux, and Windows
# Clone the repository
git clone https://github.com/openteams-lab/openteams.git
cd openteams
pnpm i
npm run dev
# build
pnpm run desktop:build
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
Windows troubleshooting (optional)
On Windows, if pnpm run dev reports error: no such command: watch, install cargo-watch and run the development command again. This installation is only needed when that error occurs.
cargo install cargo-watch
Contributing
Contributions are welcome, including reusable AI team workflows that others can learn from and adapt. Here is how to get started:
- Find an issue — Check Good First Issues for beginner-friendly tasks, or browse open issues.
- Discuss before building — Before opening a large pull request, please open an issue or discussion so the direction can be aligned.
- Follow the code style — Run the following before submitting:
pnpm run format
pnpm run check
pnpm run lint
- Submit a PR — Describe what you changed and why. Link the related issue if applicable.
See CONTRIBUTING.md for the full guide.
License
openteams is released under the Apache License 2.0. In practical terms, you may:
- use it free of charge for personal, educational, internal, or commercial projects;
- copy, modify, and build on the source code;
- distribute the original or a modified version, as source code or compiled software;
- include it in a proprietary product and charge for that product without having to open-source the rest of your code.
If you redistribute openteams or a modified version, include a copy of the license, keep the relevant copyright and attribution notices, and clearly state which files you changed.
Three other points are worth knowing:
- Brand: You may use the code, but you may not present yourself as the official openteams project or use its name and trademarks as your own brand.
- Patents: Contributors give you permission to use the patents necessarily covered by their code, so they cannot use those patents to stop you from using openteams. In return, if you file a lawsuit claiming that openteams infringes your patent, you lose this patent protection. Only the patent permission ends—not your ordinary copyright permission to use the code. Most users who are not involved in patent litigation are unaffected.
- Risk: The software is provided for free as it is. You decide whether it fits your needs and accept the risks of using it; the project does not provide a warranty or compensation for problems.
This section is a plain-language summary. The LICENSE file contains the authoritative legal terms.
See LICENSE for the complete legal terms.