Key Concepts

A shared vocabulary for the whole product. These are the building blocks every other section refers to — useful for new users and for agents reasoning about the workspace.

The hierarchy

  • Workspace — the top-level container. Everything (people, teams, tasks, projects, docs, agents, channels) belongs to one workspace. You can be a member of several workspaces and switch between them.
  • Team — a unit inside a workspace (for example Backend, Frontend, Design). Each team has its own board with its own workflow states. A task always belongs to exactly one team.
  • Member — a person in the workspace, with a role that controls what they can do. People are invited by email.

Work items

  • Task — the core unit of work. It has a key (e.g. BE-14), a title and description, a status, a priority, an assignee, labels, an estimate, and relations. Lives on a team's board.
  • Subtask — a task linked to a parent task. Subtasks are ordinary tasks; the parent shows done/total progress.
  • Dependency — a "blocked by / blocks" link between tasks. When a blocker closes, the blocked tasks update in real time.
  • Project — a cross-team initiative that groups tasks toward one goal, with its own state, health, and status-update timeline.
  • Milestone — a checkpoint inside a project, usually with a target date. A task can belong to one milestone.

Conversation surfaces

  • Discussion thread — a chat attached to something (a task, a project, a brainstorm) or a standalone channel. You post messages; you can @mention people and agents.
  • Channel — a standalone discussion in the workspace (team chat). A channel can also be shared with a trusted partner workspace.
  • Brainstorm — a dedicated space to develop an idea with an agent before it becomes tasks or a project.

Knowledge

  • Knowledge base — one Markdown engine that hosts every kind of content, organized as a page tree. The kind of content is the scope:
    • doc — structured documentation (the Docs tree).
    • note — lighter, free-form pages (the Notes tree).
    • learnings — reusable patterns agents capture from past work, browsed in Context together with each learning's track record.
    • artifacts — the documents workflows produce (plans, results, learnings), attached to the task they were made for.
  • Visibility — who can read a page: private (members only), trusted (members of workspaces you trust), or public (anyone with the link).

AI

  • Agent — an AI participant. Server-side agents are configured in the workspace and triggered with @handle; local agents connect from your machine over MCP. Both act with the permissions of the person who triggered them.
  • MCP (Model Context Protocol) — the open standard that lets a local AI client connect to the workspace and operate it through a defined set of tools.
  • Workflow — a workspace-defined process an agent follows (see AI Workflows): phases and steps for planning, implementing, creating tasks — or any process of your own. Connected clients receive workflows as wf_* tools and slash commands.
  • Artifact — a document a workflow produces from a template (a plan, a result, a learning, or a kind you define), attached to the task it was made for.
  • Learning — a reusable pattern captured from past work; its usage record (applied / helped / misled) is visible in Context.
  • Run — one execution of an agent in response to an @mention: it reads context, may call tools, and posts a reply.