Skip to main content

Visual glyph review

Use this workflow when an agent needs a visual proof of selected or named glyphs before judging outlines, spacing, sidebearings, anchors, guides, or component composition.

The render_glyph_review_image tool is read-only. It renders an in-memory PNG response through MCP image content and does not write files, add guides, edit glyph data, or save the font.

When to use it​

  • After an outline or spacing proposal, before calling a mutating tool.
  • After a dry run, to give a vision-capable model a reviewable proof image.
  • When numeric path JSON is not enough to judge visual balance.
  • When components, anchors, guides, or selected nodes need visible context.

Inputs​

The tool uses the current selection when glyph_names is omitted. Pass names explicitly when reviewing a stable list.

Common overlays:

  • metrics: baseline, x-height, cap height, ascender, and descender.
  • sidebearings: origin and advance-width lines.
  • bounds: layer bounds.
  • nodes and handles: outline point review.
  • anchors: anchor crosses.
  • guides: existing layer guides.

Sample tool calls​

Review the current Glyphs selection with default overlays:

{
  "font_index": 0
}

Review named glyphs with outline details:

{
  "font_index": 0,
  "glyph_names": ["A", "B", "O"],
  "columns": 3,
  "image_width": 1600,
  "include_components": true,
  "overlays": ["metrics", "sidebearings", "bounds", "nodes", "handles", "anchors", "guides"]
}

Request a base64 fallback in metadata for clients that do not surface MCP image content:

{
  "font_index": 0,
  "glyph_names": ["n", "o", "p"],
  "include_base64": true
}

Safe prompt template​

Render a visual review image for my current selection.
Use render_glyph_review_image with default overlays first.
Do not mutate or save the font.
Then inspect the image and report only concrete visual concerns and suggested next checks.