Agent Teams AI

Agent Teams AI is a free, open-source desktop application (AGPL-3.0) that lets developers orchestrate teams of AI coding agents across Claude Code, Codex, OpenCode, Cursor, SuperGrok, GitHub Copilot, Z.AI, MiniMax, Kiro, and many more providers. The app ships with a built-in free model that requires no authentication, account, or API key to start. It runs locally on Windows, macOS, and Linux, keeping all project code on the user's machine with no Agent Teams cloud backend.

What It Is

Agent Teams AI is a multi-agent orchestration layer presented as an all-in-one desktop workspace. Users assign a high-level goal to a team; the agents then break it into tasks, assign roles (Planner, Lead, Developer, Reviewer), work in parallel, message each other, and update a real-time kanban board — all without manual task management. The user can monitor progress, drop comments, send direct messages to any agent, or approve individual tool actions depending on the autonomy level configured. The project is built with Electron 40, React 19, TypeScript 5, Tailwind CSS 3, and Zustand 4.

Core Workspace Features

The app bundles several tools that would otherwise require separate products:

• Kanban board — 5-column real-time board that updates as agents move tasks through states; supports drag, assign, and review
• Code review — diff view per task with file-level and hunk-level accept, reject, and comment controls, similar to Cursor's review flow
• Built-in terminal — visual PTY terminal scoped to each team and project, with local shell tabs, persistent history, and autocomplete
• Built-in code editor — edit project files with Git support without leaving the app
• Token analytics and budgets — tracks input, output, cache, and reasoning token usage across teams, agents, tasks, models, and runs; supports monthly token or estimated-cost budgets with alerts at 80% and 100%
• Organizations and global overview — nested org structure (departments, squads) with a live map showing team status, task progress, dependencies, and cross-team communication

Multi-Provider Orchestration

The app auto-detects installed Claude Code, Codex, and OpenCode runtimes and can connect Cursor, SuperGrok, GitHub Copilot, Z.AI, MiniMax, and Kiro from the UI. The project page states support for 75+ providers and 200+ models. Agents from different providers can be mixed within the same team, and each teammate can optionally run in its own git worktree to avoid conflicts. A Solo Mode lets a single agent self-manage tasks before expanding to a full team.

Setup Path and Autonomy Controls

No prerequisites are required beyond downloading the platform-appropriate binary (.dmg, .exe, .AppImage, .deb, .rpm, or .pacman). The free model with no auth is available immediately on first launch. Users can configure autonomy per agent: fully autonomous runs or per-action approval with notifications. The app includes a nudge system that can automatically prompt a stalled agent to continue after a known rate-limit cooldown, guarded and rate-limited to avoid spamming. The README notes 29 interface languages and a configurable notification system for task completion, agent responses, comments, and errors.

Update: v2.12.0

The latest release is v2.12.0, published August 1, 2026, with the repository last pushed on September 2, 2026. Active roadmap items include launching 24/7 autonomous teams in the cloud, automatic account switching, and handling very complex long tasks without losing context. The project's stated vision is a minimal phone-accessible interface where a lead agent routes voice instructions to the right team, with technical coordination surfaced only when human input is genuinely required. The GitHub repository shows 2,019 stars and 339 forks as of the data snapshot.