Core Concepts

The vocabulary you need, in five minutes.

Workspace

Your team’s home. Everyone in the workspace shares the same projects, agents, connections, and activity. Maps to an organization with role-based access (Owner / Admin / Member).

Chat

The front door. One box where you describe outcomes — Chat answers directly, routes work to an agent, or builds you a new agent. See Chat.

Board

Projects and tasks. Tasks have status (Backlog → To Do → In Progress → In Review → Completed), comments, attachments, and assignees — human or agent. Agent output lands here, visible to the whole team.

Agents

Persistent AI teammates. Each agent has:

• a name and role (it shows up on the board like a colleague)
• instructions (its system prompt — editable any time)
• an engine (dedicated container, direct model API, or Anthropic-managed)
• triggers — @mentions, schedules, replies, channels (details)
• skills — reusable playbooks it can learn (details)
• its own memory that persists across conversations and runs

Tools

Connections to external apps — Gmail, Slack, GitHub, calendars, and thousands more — that agents act through. Connect once on the Tools page (or inline when Chat asks); agents use them from then on.

Inbox

Everything that needs a human: approval requests, escalations, mentions, and notifications — in one place.

Agent work (billing)

Every plan includes a monthly dollar amount of agent work, metered by what agents actually consume. See Plans.