generate_pages_preview Leave feedback

generate_pages_preview renders pages as PNG images returned inline — with annotations baked in — so a client that supports image content can show them, and the agent can look at what it just did. Pass pages as 1,3,5 or 1-3; omit it for page 1. Example prompt: “Show me page 2 with the highlight”.

Tool description (as the AI agent sees it):

Renders document pages as PNG images and returns them inline so AI clients can display them directly. Supports PDF, DOCX, XLSX, PPTX, and 50+ more document and image formats. Pass ‘pages’ as a comma-separated list (e.g. ‘1,3,5’) or a range (‘1-3’); omit to render only page 1. Renders the document with its annotations baked in — call this whenever the user asks to preview, show, render, or visualise a page (annotated or not). Maximum MaxPagesPerCall (5) pages per call; further pages must be requested in additional calls. Do NOT pre-check whether files exist — pass the filename the user provided directly. Returns a CallToolResult with one TextContentBlock describing what was rendered, followed by one ImageContentBlock per page. On failure, the response contains a single TextContentBlock starting with ‘Preview generation failed for’.

{
  "name": "generate_pages_preview",
  "arguments": {
    "file": {
      "filePath": "contract_annotated.pdf"
    },
    "pages": "1-3"
  }
}

One text block describing what was rendered, followed by one image block per page.

This is the tool that closes the loop on annotation work: the agent adds a highlight, renders the page, and can verify the placement rather than asserting it. Rendering is the most expensive call in this server — ask for the pages you need, not the whole document.

On failure a single text block starts with Preview generation failed for.

• “Show me page 2 with the highlight you just added.”
• “Render pages 1-3 so I can check the annotations.”
• “Preview the first page of the annotated contract.”

See it used end-to-end: Let the agent see the page.