get_image_data

Downloads the bytes of an image and returns them as MCP image content, so a multimodal LLM can look at what's on the canvas. Identify the image by itemId (an image item on the canvas — its file id is resolved for you) or by a raw fileId.

Tool name: get_image_data

Parameters

Parameter	Type	Required	Default	Description
projectId	string	Yes	—	Project (map) id the image belongs to.
itemId	string	One of these	—	Image item id; its stored file id is resolved automatically.
fileId	string	One of these	—	Stored file id (alternative to itemId). get_project lists a fileId for image items.
thumbnail	boolean	No	false	When true, fetch the smaller preview instead of the original.

Exactly one of itemId or fileId must be provided.

Returns

An MCP image content block — base64 data plus the MIME type reported by the file store:

{
  "content": [
    { "type": "image", "data": "<base64>", "mimeType": "image/png" }
  ]
}

Example

"What does the diagram on the board show?"

{
  "projectId": "a1b2c3d4-...",
  "itemId": "33cc44dd-...",
  "thumbnail": true
}