# Canvas

> A map of one agent — what is bound to it, drawn as a graph you can click into.

The **canvas** draws an agent and everything attached to it: the skills and connectors it can use, the channels it answers on, the workspace it owns. It is the same configuration as the [Agent Builder](/docs/build/agent-builder/) sections, seen all at once.

<Screenshot
  name="studio-canvas"
  alt="The canvas: the agent as a hexagon in the middle showing its default model, with API, MCP, device, sub-agent, channel and connector axes drawn around it and the bound Slack, Gmail and Notion connections on the ends."
  caption="An axis with nothing on it still appears — the gaps are as visible as the connections."
/>

## When to use it

Open the canvas when you want to know *what this agent actually is* rather than change one setting — before handing it a channel, when a colleague asks what it can reach, or when an agent behaves in a way its prompt does not explain.

Use the sections when you know what you are changing. The canvas is for looking.

## Reading it

The agent sits in the middle. Around it are the things bound to it, grouped by kind — abilities, pairings, inventory. An axis with nothing on it still appears, so the gaps are as visible as the connections.

Selecting a node opens the same panel the corresponding section uses, so you can go from *"what does this agent have"* to changing it without leaving the map.

<Aside type="note" title="Some axes are shown before they are configurable">
  A few nodes exist to show the shape of the model rather than to be edited yet.
  If a node opens a read-only panel, that axis is not configurable from here.
</Aside>

## Limits

- The canvas shows one agent. There is no view across all your agents.
- It reflects configuration, not activity — what the agent *can* reach, not what it used today. For that, see [usage](/docs/account/plans-and-usage/) or the [run](/docs/work/runs/) itself.
- Editing still happens in panels. The canvas is a way in, not a separate editor.

## Related

<CardGrid>
  <LinkCard
    title="Agent Builder"
    href="/docs/build/agent-builder/"
    description="The same configuration, section by section."
  />
  <LinkCard
    title="Security and permissions"
    href="/docs/account/security/"
    description="What being bound actually grants."
  />
</CardGrid>