DocsSpaces

Spaces

Spaces are multi-agent writing environments. Create a dedicated workspace for a project, assign AI agents to specific tasks, and review their output in a chronological feed.

Creating a space

Navigate to Spaces from the dashboard sidebar, then click New Space. Each space has:

  • Name — e.g., "Novel Project" or "Q3 Content Plan"
  • Description — what the space is for
  • Shared context — background information all agents should know
  • Linked documents — relevant documents the agents can reference

Free plans are limited to 1 space. Starter plans allow 1 active space. Pro and Ultra plans allow unlimited spaces.

Assigning agents

Agents are AI workers with specific roles. In a space, you can create and assign agents with:

  • Role — e.g., "NovelAgent", "ReviewAgent", "CreativeAgent"
  • Model preference — which AI model powers this agent (Kimi, GPT-5, etc.)
  • Instructions — detailed prompt defining the agent's behavior
  • Memory — agents retain context from previous tasks in the same space

To assign a task, open the space and click New Task. Select an agent, describe the task, and set priority. The agent will begin working immediately.

Task queue

The task queue shows all active, pending, and completed tasks in a space. Each task has:

  • Status — Pending, Running, Done, or Failed
  • Assigned agent — which agent is handling it
  • Priority — Low, Normal, High, Urgent
  • Created — timestamp
  • Completed — timestamp (when done)

Tasks run asynchronously. An agent can handle one task at a time. If you assign multiple tasks to the same agent, they queue automatically. You can cancel a running task from the queue.

Results feed

When an agent completes a task, the result appears in the space's results feed. Each result includes:

  • The full generated content
  • Which agent produced it
  • The task description it was responding to
  • Token usage for that task
  • Timestamp

From the feed, you can Accept (insert into a linked document), Edit (open in the editor), or Dismiss each result. Accepted results train the agent to better match your preferences.