AI teammates are specialists with a job and clear limits

AI teammates are part of the organization's AI workforce. Each one is set up for a role, such as answering customer questions, drafting documents, monitoring operations, or helping a team complete repeatable work.

Think of an AI teammate like a specialist on your team. The teammate has a job, a set of approved capabilities, and boundaries that define what the teammate can access or change.

Teams manage people and AI teammates together in Staff. From there, they can add an AI teammate, open the detail page, chat with the teammate, review logs, and manage configuration.

An AI teammate detail page can include rules, skills, storage access, MCP server access, logs, reports, jobs, schedulers, and API setup depending on how that teammate is configured.

Teams can configure the AI teammate's description, role, instructions, model, tools, voice settings, skills, storage access, and connected MCP servers.

This makes AI teammates flexible without making them unlimited. The useful pattern is to give each teammate enough context and capability for the role, not every permission in the workspace.

People can chat with AI teammates from the internal chat surface, mention teammates when supported, attach files, use voice when enabled, and review conversation history.

When a task needs human judgment, Alloy keeps people in the flow. AI teammates are meant to work with the team, not replace the team's control over the work.

Non-system AI teammates only see storage folders explicitly shared with them. They can also use approved tools, assigned MCP servers, and skills connected to their role.

This lets teams separate responsibilities. A support teammate can have customer-facing tools, while an operations teammate can have internal workflow tools.