agentdraw.app

inspect_canvas

Read every element on the whiteboard in a compact form: type, bounding box, text or label, bindings, absolute points for strokes and lines (sampled to 64), and for the person’s strokes the nearIds of shapes they touch. Also returns the viewport (scene coordinates), what the person has selected, the scene bounds, a suggested empty origin for new content, and how many elements the person changed since you last looked. The only type you can read but not create is freedraw. Embeddables carrying html report html: { chars, excerpt } instead of a link; pass includeHtml: true together with ids to get html.text, the full source. Pass sinceRevision to get only the elements that changed since then, plus the change log. detail: "outline" drops styles and points and keeps geometry, text, and bindings; detail: "graph" returns the board as a graph (elements are the nodes, plus graph.edges with from/to/text, graph.frames with their children, and graph.marks for the person’s strokes), the cheapest way to understand a flow or architecture diagram; every element carries by: "agent" or "person"; limit (default 250, max 500) and cursor page through large boards, and nextCursor tells you when there is more; a cursor is bound to the revision it was issued at and answers cursor_expired if the board changed in between, so start over from the first page. A label is a separate text element bound to its shape, so a labelled shape counts as two elements everywhere: elementCount counts the whole board, matchedCount what matched after filters, returnedCount what this page holds. Writes need the epoch and revision from this call or from your last apply_patch response.

Input schema

{
  "type": "object",
  "properties": {
    "sinceRevision": {
      "type": "integer",
      "minimum": 0
    },
    "ids": {
      "type": "array",
      "minItems": 1,
      "maxItems": 200,
      "uniqueItems": true,
      "items": {
        "type": "string",
        "pattern": "^[A-Za-z0-9][A-Za-z0-9._-]{0,119}$"
      }
    },
    "frameId": {
      "type": "string",
      "pattern": "^[A-Za-z0-9][A-Za-z0-9._-]{0,119}$"
    },
    "viewportOnly": {
      "type": "boolean"
    },
    "detail": {
      "type": "string",
      "enum": [
        "full",
        "outline",
        "graph"
      ]
    },
    "includeHtml": {
      "type": "boolean"
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 500
    },
    "cursor": {
      "type": "string",
      "pattern": "^c[0-9]{1,7}\\.[0-9]{1,9}$"
    }
  },
  "additionalProperties": false
}