Add your first AI teammate

AI teammates live in the Staff area next to human team members. The main Staff page shows both people and AI teammates in one directory.

To create one, use Add AI Teammate. The create route is /staff/ai/new. In create mode, Alloy starts with the Rules tab so you can define the teammate before configuring the full setup.

Start with the role. A good AI teammate has a specific responsibility, such as customer support triage, document drafting, internal onboarding, operations follow-up, or API-based task handling.

Use the rules and instructions to say what the teammate does, what the teammate does not do, when the teammate asks for help, and what a good result looks like.

After the teammate is created, the full setup can include tabs such as Rules, Skills, Storage, MCPs, Logs, Reports, Jobs, and Schedulers.

The exact setup depends on the job. Some teammates only need instructions and shared knowledge. Others need MCP servers, skills, scheduled jobs, reports, or logs for monitoring.

AI teammates can be configured for different channels. Current header choices include Not selected, Web Chat, and API.

Use Web Chat when the teammate needs to be available through an embedded chat experience. Use API when another system calls the teammate programmatically.

The AI teammate Storage tab shows folders explicitly shared with that teammate. The teammate does not automatically see every file in the workspace.

Use Storage sharing to give the teammate read or write access to the folders the teammate needs. Keep access scoped to the job.

The MCPs tab lets users enable or disable organization MCP servers for a teammate. MCPs and other tools are useful when the teammate needs to work with outside systems or perform approved actions.

Do not start by connecting everything. Start with the smallest set of tools and knowledge that lets the teammate do the job.