Agent Teams

Agent Teams is a free desktop application that orchestrates teams of AI agents working on coding and project tasks. Users assemble a team with different roles that work autonomously in parallel, watch progress on a kanban board, and give high-level commands rather than micromanaging each task. It targets developers and teams who want to run multiple coding agents, such as Claude Code, Codex, and Cursor, together on larger projects.

Agents in a team communicate with each other, create and manage their own tasks, review each other's work, and leave comments, while task status changes automatically on the kanban board. Cross-team communication lets agents message or collaborate across different teams. Users can send a direct message to any agent, comment on a task, or pick a quick action on a kanban card at any time, and can let agents run fully autonomous or require approval before each tool action.

Agent Teams tracks token usage and cost across teams, agents, tasks, and runs, with configurable monthly budgets and alerts at 80% and 100%. It groups teams into organizations with a live map of status and dependencies, supports a solo one-agent mode to save tokens, includes an integrated terminal workspace per project, and offers 29 interface languages. A built-in code editor with Git support and a visual diff review let users inspect and approve agent changes like in Cursor.

Agent Teams has no prerequisites and can detect installed Claude Code, Codex, and OpenCode runtimes, or connect Cursor, SuperGrok, GitHub Copilot, Z.AI, MiniMax, and Kiro from its interface. It is released under the AGPL-3.0 license and distributed as a free desktop app, with no signup, API key, or card needed to start with its free model. It supports 300+ models and 200+ LLM providers and runs as an Electron desktop application.

Use cases

• Running large or complex projects that need bigger teams of coordinated agents
• Getting full visibility into what each agent is doing across teams