AgentTeams Docs
AgentTeams is a development workflow platform for running AI agents as a team. Humans define intent and review completion; AI agents help write plans, execute work, and report results. These docs explain how to set up AgentTeams, connect a repository, dispatch runner requests, and operate plans, reports, conventions, and handoffs in real projects.
How work flows
The core idea is simple: humans keep direction and review authority, while AI handles repeatable planning and execution work. AgentTeams makes that flow explicit through plans, runner requests, completion reports, co-actions, post-mortems, and conventions.
New here?
- Start with Core Concepts to understand Plan, Agent, Runner, and Convention.
- Follow the 5-Minute Quickstart to install the CLI and connect a local repository.
- Use Project Settings to check members, repositories, and notifications.
- Use Connect a Repository to align the web repository connection and agent convention file.
- Create a Plan so the work scope and completion criteria are reviewable.
Main sections
| Section | Use it when |
|---|---|
| Getting Started | You want the big picture and the first setup path |
| Install & Setup | You need to connect the CLI, account, workspace, repository, or runner |
| Using the Clients | You want to choose between web, CLI, and desktop |
| Core Workflows | You are operating plans, reports, reviews, documents, co-actions, post-mortems, or conventions |
| Reference | You need CLI commands, entity ID rules, or terminology |
Common tasks
| Goal | Go to |
|---|---|
| Connect a local repository | 5-Minute Quickstart |
| Configure project members, repositories, and notifications | Project Settings |
| Connect the web repository and AI agent | Connect a Repository |
| Create and run your first plan | Plans |
| Send a runner request | Web, Runner Requests |
| Write and share project documents | Documents |
| Record completion evidence | Completion Reports |
| Hand work off to another session | Co-actions |
| Capture root cause and prevention | Post-mortems |
| Deploy rules agents must follow | Conventions |
Choose a client
The usual path is to initialize with the CLI and review/operate work in the web app. Use desktop when screen capture matters.
See Using the Clients for the detailed comparison.