Agent Teams DocsRun AI agent teams from a local desktop app

Create teams, watch work move across a kanban board, review code changes, and coordinate Claude, Codex, OpenCode, and multimodel workflows without giving up local control.

Beginner workflow

Install

Concepts

Team-first workflow

Define roles, launch a lead, and let agents split, claim, and coordinate tasks.

Create a team

Live kanban board

Watch tasks move through todo, in progress, review, done, and approved as agents work.

Understand workflow

Built-in code review

Inspect task-scoped diffs, accept or reject hunks, and comment where agents need direction.

Review changes

Runtime-aware setup

Use Claude, Codex, OpenCode, or multimodel providers through the access you already have.

Configure runtimes

Local-first control

The desktop app reads local project and runtime state. Your code stays on your machine unless a selected provider receives prompt context.

Privacy model

Debuggable teams

Trace task logs, runtime output, teammate messages, and live processes when a launch or task gets stuck.

Troubleshoot

Start here

Agent Teams is a free desktop app for orchestrating AI agent teams. You are not just sending isolated prompts to one agent: you create a team, assign roles, and watch agents coordinate work through a task board.

01Beginner workflowUnderstand the first run from project to approval.02QuickstartInstall the app and validate the base launch.03First teamLead, builder, reviewer, roles, models, and Worktree.04Run workLead brief, task board, comments, and monitoring.05Review and approveTask detail, logs, diff, and hunk-level decisions.06Runtime setupClaude, Codex, OpenCode, and multimodel setup.

Next steps after launch

After creating your first team, explore these guides to go further:

Runtime setup - configure Claude, Codex, OpenCode, or multimodel providers: Configure runtimes
Beginner workflow - follow the complete first-run path from project to approval: Start the walkthrough
Create your first team - set up lead, builder, reviewer, roles, models, and Worktree: Create the team
Run and monitor work - read the board, comments, task detail, and logs: Run the team
Review and approve - inspect task results and code changes before approval: Review work
Agent workflow - understand how agents coordinate through the task board: Understand workflow
Team brief examples - learn prompt patterns from real-world briefs: See examples
Code review - inspect diffs, accept or reject changes: Review changes
Troubleshooting - diagnose stuck launches, missing teammates, and task failures: Troubleshoot
Git worktree strategy - use worktree isolation when multiple teammates edit the same repo in parallel: Learn about worktrees
Release notes - see what's new in each version: View releases

Reference

Use the reference pages when you need exact terminology, provider behavior, contributor architecture, or privacy boundaries.

◈ConceptsTeams, tasks, roles, and autonomy levels.⌁RuntimesClaude, Codex, OpenCode, and multimodel mode.▦ArchitectureFeature layout, guardrails, and runtime/provider boundaries.⌘Local dataWhat stays on disk and what providers receive.?FAQShort answers to common questions.

Agent Teams DocsRun AI agent teams from a local desktop app

Team-first workflow

Live kanban board

Built-in code review

Runtime-aware setup

Local-first control

Debuggable teams

Start here ​

Next steps after launch ​

Reference ​

Product preview ​

Start here

Next steps after launch

Reference

Product preview