glyphs-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@glyphs-mcp Audit the color of my lowercase letters"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP bridge for AI-assisted type design in GlyphsApp.
Lets Claude, Cursor, or any MCP client read and write font data directly in GlyphsApp — bidirectional, real-time, live in the editor.
MCP Client ←(stdio/MCP)→ MCP Server ←(HTTP/localhost)→ GlyphsApp PluginRequirements
An MCP client
Python 3.10+ with uv (recommended) or pip
glyphsetsandshaperglot(installed automatically with the MCP package)
Related MCP server: FontLab MCP Server
Installation
1. Install the GlyphsApp plugin
From Plugin Manager (recommended): Open GlyphsApp, go to Window > Plugin Manager, search for MCP, and click Install.
Manual install: Download GlyphsMCP.glyphsPlugin.zip from the latest release, unzip, and double-click to install.
Restart GlyphsApp. You should see GlyphsMCP under the Window menu.
2. Connect your MCP client
Open Window > GlyphsMCP > Connect and choose your client:
Claude Code — registers a user-scoped stdio server with the official
claudeCLI.Codex / ChatGPT Desktop — registers the server with
codex; both apps share this configuration.OpenCode — adds the server to OpenCode's global configuration; restart OpenCode afterward.
Visual Studio Code — adds the server to your active user profile with
code --add-mcp.Cursor — safely merges the server into
~/.cursor/mcp.jsonand backs up an existing file first.Other MCP Client — copies a portable
mcpServersJSON block to the clipboard.
Glyphs 4 registers as glyphs-mcp and Glyphs 3 as glyphs-mcp-3, each using the port configured in that app, so both can stay registered without collision. If a client CLI is not available to GlyphsApp, the equivalent Terminal command is copied to the clipboard instead.
For manual setup, use the endpoint shown when starting the GlyphsMCP server. For example:
{
"mcpServers": {
"glyphs-mcp": {
"command": "uvx",
"args": ["glyphs-mcp"],
"env": {
"GLYPHS_URL": "http://127.0.0.1:7746"
}
}
}
}Claude Desktop now recommends Desktop Extensions (.mcpb) for local MCP servers. A GlyphsMCP Desktop Extension is planned separately and is not installed by this menu yet.
3. Use it
Open a font in GlyphsApp, then ask your AI assistant:
"Run a full color audit on my font and tell me which glyphs are inconsistent"
"Compare the stems across all my lowercase letters"
"Check if my figures are consistent with my letters"
"Make the R 20% wider and harmonize the curves"
"Analize metrics on lowercases"
Tools
Read
Tool | Description |
| Font family name, UPM, glyph count, masters, axes, metrics, instances |
| Glyph metadata, optionally filtered by |
| Full glyph data: paths, components, anchors, sidebearings for all layers |
| Glyph rendered as SVG markup |
| Current editor selection: active glyph, layer, selected paths/nodes |
| All masters with metrics and axis positions |
| Kerning pairs filtered by |
| OpenType feature code |
Write
Tool | Description |
| Create a new glyph with optional width, unicode, and initial paths |
| Replace all paths on a glyph's layer |
| Set advance width |
| Set color label (0-11) |
| Assign or clear a unicode value |
| Rename a glyph |
| Copy a glyph with all layers to a new name |
| Delete a glyph |
| Add or modify a kerning pair |
| Remove a kerning pair |
| Create or update an OpenType feature |
| Generate U+2500–U+259F outlines across masters; can overwrite drawings |
| Export OTF, TTF, WOFF, WOFF2, and variable TTF beside the saved source |
Analysis
Tool | Description |
| Measure stem thicknesses via perpendicular ray-casting |
| Compare stems across glyphs using industry patterns |
| Designer's intended stem values from Dimensions palette |
| Measure ink density for a single glyph |
| Compare ink density across glyphs |
| Full font color audit across all letters |
| Overshoot consistency at baseline, x-height, cap-height |
| Width ratios, related-form groups, ordering constraints |
| Diagonal stem thickness vs straight reference |
| Stem thinning at arch/bowl junctions |
| Cross-validate figures and letters (0/O, 6/9, 8/S, 3/B) |
| Mirrored pair widths, dash ratios, related punctuation |
| Master compatibility: paths, nodes, components, anchors |
| Kerning quality: cross-master gaps, orphans, outliers |
| Spacing quality: sidebearing groups, symmetry, drift |
| Analyze or assign kerning groups, with dry-run/overwrite control |
| Preview or apply optical-area kerning to critical, automatic, or explicit pairs |
| Check Google Fonts glyphsets and optionally add missing empty glyphs |
| Export a temporary instance and evaluate language support with Shaperglot |
| Run a 44-item production-readiness review |
| Screen a proposed family name against Fontdata; not legal clearance |
Analysis tools never change glyph colors by default. Pass mark_glyphs=True only when you explicitly want verdict labels in GlyphsApp: red = inconsistent, orange = unreliable, yellow = warning/optical compensation, green = pass. analyze_kerning, analyze_spacing, check_compatibility, and compare_stems also point to matching workflow recipes for broader checks.
RMX Tools
Requires RMX Tools for full functionality. rmx_scale now uses real RMX processing, accepts one value or a per-master list for scale parameters, and defaults allow_fallback=False; native affine fallback must be explicitly enabled. rmx_tune delegates to the loaded RMX Tuner and supports blend and all_masters.
Tool | Description |
| Optimize bezier curves |
| Scale by percentage with stroke weight compensation |
| Adjust weight, width, height, or slant |
| Adjust a glyph to a fixed advance width |
| Apply any RMX filter to multiple glyphs |
| Scale multiple glyphs with measured stem compensation, optional backups |
Recipes
Bundled markdown recipes provide ordered workflows for consistency audits, spacing, kerning, master compatibility, proportional scaling, dated-layer cleanup, and Glyphs plugin or script creation.
Tool | Description |
| List bundled and user-created recipes |
| Read a complete recipe |
| Read one numbered step with its next-step directive |
| Write a recipe markdown file; overwrite is opt-in |
| Permanently delete a recipe markdown file |
Recipe creation and deletion modify files in the installed plugin's Resources/recipes directory.
Recipes are also exposed as MCP prompts, so clients that surface prompts as commands get a one-command entry point. The prompt name is the recipe name. With the server registered as glyphs-mcp, a recipe runs as /glyphs-mcp:<recipe> or /mcp__glyphs-mcp__<recipe>:
Recipe | Command (Claude Code) | What it does |
|
| Full font consistency audit: stems, color, proportions, spacing |
|
| Systematic spacing pass following Cheng/Briem/Ruder |
|
| Kerning from scratch: groups, critical pairs, verification |
|
| Scale glyphs with automatic stem-weight compensation |
|
| Pre-export master compatibility and metrics check |
|
| Safely remove timestamped backup layers |
|
| Create a GlyphsApp plugin bundle |
|
| Create a GlyphsApp Script-menu script |
|
| List the available recipes, or start one by name |
These forms are what Claude Code uses. On Glyphs 3 the server prefix is glyphs-mcp-3 (for example /glyphs-mcp-3:audit_consistency), so both apps can be registered side by side. MCP-prompt support varies by client: if yours does not surface prompts as commands, ask for the recipe by name or use list_recipes / get_recipe / get_recipe_step. In opencode you can add a custom command under .opencode/commands/ (or ~/.config/opencode/commands/) that calls get_recipe_step to get a literal /create-plugin.
Advanced
Tool | Description |
| Run arbitrary Python inside GlyphsApp (disabled by default) |
Multi-master support
All tools accept an optional master_id parameter. When omitted, read/write tools use the first master. Analysis tools analyze all masters and return per-master results.
Tools with side effects are explicit: kerning-group analysis can assign groups, while audit color labels require mark_glyphs=True; auto-kern can write kerning; glyphset coverage can add empty blue-labelled glyphs; box drawing creates or replaces outlines; smart scale modifies outlines and can create backup layers; recipe CRUD writes or deletes markdown files. export_font writes a new timestamped directory under export/ beside the saved source and never removes previous exports. Use dry-run/preview options where available and save the font before bulk operations.
Font export
Ask the agent to export the open font, or run the bundled CLI directly:
uvx --from glyphs-mcp export-glyphs /path/to/MyFont.glyphsBoth paths use the official glyphs-cli, installed automatically with the GlyphsMCP server package on macOS. Glyphs Plugin Manager installs only the app plugin; uvx glyphs-mcp provisions the server and export CLI on first use. Each run creates export/YYYY-MM-DD_HH-MM-SS/ beside the source, with separate otf, ttf, woff, woff2, and variable directories as applicable, plus export-report.jsonl. The MCP tool refuses to export unsaved changes unless save_before_export=True is explicitly requested. Every result includes portable exportLog and ANSI-colored exportLogAnsi variants; the calling agent must show the colored variant when its client supports terminal colors and fall back to the portable log elsewhere.
Menu
The plugin adds a GlyphsMCP submenu under Window in the menu bar:
Start/Stop Server — toggle the HTTP server
Connect — register GlyphsMCP with Claude Code, Codex/ChatGPT Desktop, OpenCode, VS Code, Cursor, or copy a generic configuration
Documentation — open this page in your browser
Allow Execute Endpoint — enable
execute_in_glyphs(off by default for security)
Preferences
Key | Default | Description |
|
| HTTP server port |
|
| Start server on GlyphsApp launch |
|
| Enable the execute endpoint |
How it works
The GlyphsApp plugin runs an HTTP server on 127.0.0.1 using the configured port (7745 by default). All GlyphsApp API calls run on the main thread via a queue + NSTimer bridge for thread safety.
The MCP server is a thin translation layer — it receives MCP tool calls via stdio and forwards them to the plugin URL supplied through GLYPHS_URL.
Bundled agent skill
Python distributions include the complete skills/type-designer/ skill: a concise runtime contract plus references for general type-design workflows, typeface DNA, Glyphs 4 and scripting, safe GlyphsMCP operation, the audit contract, and curve-continuity evidence. GlyphsMCP is one execution environment within the skill, not its only subject.
The skill is not installed automatically or served live through SEP-2640 because the current Python MCP SDK does not expose that extension. Install the directory through your client's normal skill workflow. To keep this checkout as the single source of truth, link rather than copy it into each client's discovery directory:
ln -s "/path/to/glyphs-mcp/skills/type-designer" "$HOME/.codex/skills/type-designer"
ln -s "/path/to/glyphs-mcp/skills/type-designer" "$HOME/.claude/skills/type-designer"
ln -s "/path/to/glyphs-mcp/skills/type-designer" "$HOME/.config/opencode/skills/type-designer"Every client then discovers the same physical skill through SKILL.md; no separate copies diverge.
Roadmap
Font proofing — Generate proof strings for spacing/kerning evaluation
Auto-update — Check for updates directly from the GlyphsMCP menu
Analytics — Optional usage telemetry to guide development priorities
License
MIT — Nicolas Massi www.nico.works
Available Tools
55 toolsanalyze_kerningA
Analyze kerning quality across all masters.
RECIPE: For kerning from scratch, call get_recipe("kerning_from_scratch") first — it covers group assignment, critical pairs, and verification.
Checks for:
Cross-master missing pairs (pair in some masters but not all — causes interpolation jumps)
Cross-master sign changes (positive in one master, negative in another)
Outlier values (extreme kerning > 40% of UPM)
Redundant exceptions (glyph-level overrides that match group value — can be removed)
Group orphans (Letter glyphs missing kerning group assignments)
Returns a formatted markdown report without modifying glyph colors. When
mark_glyphs=True, marks cross-master issues red and warnings yellow.
Args: master_id: Optional master ID (cross-master checks always run across all masters) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: master_id, mark_glyphs. Returns: result
analyze_kerning_groupsA
Analyze and assign kerning groups to all glyphs.
RECIPE: For a complete kerning workflow (groups → pairs → verification), call get_recipe("kerning_from_scratch") first.
Assigns correct groups to all glyphs, overwriting any existing values. Uses a five-tier resolution strategy:
Dictionary lookup for ~80 base glyphs (A-Z, a-z, figures, punctuation)
Dot-suffix stripping (a.ss01 → a)
Component inheritance for accented/composite glyphs (Aacute → A)
Unicode decomposition fallback
Contour analysis fallback (ray-casting edge detection)
Group names follow professional conventions (key glyph = group name):
UC left: H (straight stem), O (round), A (diagonal), V, T, S, etc.
UC right: H (straight), O (round), D (half-round), E (horizontal), etc.
LC left: h (straight), o (round), v (diagonal), f, etc.
LC right: h (straight), n (arch), o (round), etc.
Figures: each gets its own group (shapes too varied)
Does not modify glyph colors by default. Set mark_glyphs=True only when
the user explicitly asks to mark applied or proposed group changes.
Args: glyph_names: Optional list of glyph names (default: all Letter/Number/Punctuation glyphs) apply: If True (default), assign groups. If False, dry run only. overwrite: If True (default), overwrite existing groups. If False, only assign to empty slots. mark_glyphs: Color affected glyphs only when explicitly requested.
Inputs: apply, overwrite, glyph_names, mark_glyphs. Returns: result
analyze_spacingA
Analyze spacing quality across all masters.
RECIPE: For systematic spacing work, call get_recipe("spacing_workflow") first — it follows Cheng/Briem/Ruder methodology step by step.
Measures sidebearings and white space margins using scanline ray-casting, then checks for consistency issues:
Sidebearing group consistency (n-group: h,i,k,l,m,n,p,r should match; o-group: c,d,e,g,o,q)
Tracy/Smith per-glyph sidebearing rules (from "Designing Type" by Karen Cheng)
Side-type ordering (straight SB > round SB > diagonal SB)
Symmetric glyph check (o, O, H, I should have LSB ≈ RSB)
Reference ratios (n LSB / o LSB — optimal ~1.5x, acceptable 1.2–2.0)
Counter-based validation (n LSB should be 25–50% of n counter width)
Word space check (space width ≈ ¼ em ≈ width of i)
Cross-master spacing drift (spacing ratios should be maintained)
Does not modify glyph colors by default. When mark_glyphs=True, marks
significant inconsistencies red, minor deviations yellow, and passes green.
Args: master_id: Optional master ID (empty = all masters) glyph_names: Optional list of glyphs (empty = all Letter glyphs) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: master_id, glyph_names, mark_glyphs. Returns: result
audit_font_colorA
Full font color audit — analyzes ALL letter glyphs grouped by category.
RECIPE: For a full consistency audit, follow get_recipe("audit_consistency").
Groups glyphs into uppercase, lowercase, and figures, then evaluates each against expected density ratios from industry patterns.
If master_id is omitted, analyzes ALL masters with worst-verdict-wins for glyph colors in GlyphsApp.
Does not modify glyph colors by default. Set mark_glyphs=True only when
the user explicitly asks to mark results in GlyphsApp.
The lowercase-to-uppercase density ratio is typically 1.10-1.16 in professional fonts.
This is the comprehensive "is my font's color even?" check. Use this before final production to catch any weight inconsistencies.
Args: master_id: Optional master ID (empty = all masters) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: master_id, mark_glyphs. Returns: result
auto_kernA
Auto-kern glyph pairs using optical gap analysis (MB LetterKerner algorithm).
Measures the optical white area between each glyph pair using horizontal ray-casting, weighted by a trapezoidal function (full weight at baseline–xHeight, tapered in descender/ascender zones), then solves for the kern value that makes each pair's gap area match a calibration reference.
Auto-calibrates from nn (lowercase) and HH (uppercase) if no area is given.
Recommended workflow:
Finalize spacing first (kerning can't fix bad spacing)
Assign kerning groups: analyze_kerning_groups(apply=True)
Preview: auto_kern(pairs="critical", dry_run=True)
Apply critical: auto_kern(pairs="critical")
Expand: auto_kern(pairs="auto") — all group combinations
Verify: analyze_kerning()
Args: pairs: Pair selection mode: "critical" — ~75 essential pairs (AV, AT, To, Va, etc.) "auto" — generate representative pairs per kerning group combination "explicit" — use pairs_list pairs_list: Explicit pairs when pairs="explicit", e.g. [["A","V"],["T","o"]] area: Target gap area in units². None = auto-calibrate from nn/HH. step: Vertical sampling interval (default 5u, smaller = more precise) depth: Max probe depth per side (default 200u) factor: Optical correction factor (default 1.25, matches HT LetterSpacer) rounding: Round kern values to multiples of this (default 5) threshold: Skip kern values with abs < threshold (default 3) use_groups: Use group kerning keys (default True, strongly recommended) overwrite: Overwrite existing kerning (default False) dry_run: Preview only, don't apply (default False) master_id: Process only this master (empty = all masters)
Inputs: area, step, depth, pairs, +8 more. Returns: result
check_compatibilityA
Check master compatibility across all glyphs in the font.
RECIPE: For a complete pre-export check, call get_recipe("master_compatibility") first — it includes metrics interpolation and alignment zone verification.
Compares layers across masters for each glyph, checking:
Path count, node count, node types, path directions
Path order (spatial position must match across masters)
Starting node positions
Component count and names
Anchor names
Does not modify glyph colors by default. When mark_glyphs=True, marks:
Red (0) = incompatible (structural mismatch between masters)
Orange (1) = empty or missing drawing in one or more masters
Green (4) = fully compatible
Args: glyph_names: Optional list of glyphs to check (default: all glyphs) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: glyph_names, mark_glyphs. Returns: result
check_diagonal_weightsA
Check diagonal stroke weight consistency and ratio to vertical stems.
Measures perpendicular thickness of diagonal strokes (V, A, W, X, Y, Z, v, w, x, y, z, k, M, N) and checks:
Related diagonal groups are consistent (V≈A≈W, v≈w≈y, etc.)
Diagonal/straight ratio within professional range (typically 85-100%)
If no glyph_names provided, checks all diagonal glyphs. If master_id is omitted, analyzes ALL masters.
Does not modify glyph colors by default. When mark_glyphs=True, marks
inconsistencies red, ratio warnings yellow, and passing glyphs green.
Args: glyph_names: Optional list of glyphs to check master_id: Optional master ID (empty = all masters) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: master_id, glyph_names, mark_glyphs. Returns: result
check_font_nameA
Screen one proposed typeface family name for known collisions.
Queries the public Fontdata Namecheck API and returns exact, close, and partial match counts, trademark information, a review status, timestamp, and a link to the full result. This is collision screening, not legal clearance: a name that is not found is not guaranteed to be available.
Args: name: Proposed typeface family name to check.
Inputs: name (required).
check_glyphset_coverageA
Check font coverage against Google Fonts glyphset definitions.
With no arguments, lists all available glyphsets. With a glyphset name, reports which glyphs are present and which are missing.
Use add_missing=True to create empty glyphs for all missing characters, ready for the designer to draw.
Args: glyphset: Glyphset name or shortcut. Examples: "GF_Latin_Core" — required for Google Fonts onboarding (324 glyphs) "GF_Latin_Plus" — extended Latin (141 additional glyphs) "latin_core" — shortcut for GF_Latin_Core "latin_plus" — shortcut for GF_Latin_Core + GF_Latin_Plus (465 total) "latin_all" — shortcut for Core + Plus + Beyond (598 total) "cyrillic_core", "greek_core", "arabic_core" — other scripts Empty string — list all available glyphsets and shortcuts add_missing: If True, create empty glyphs for all missing characters in the font. New glyphs are marked with blue color label.
Inputs: glyphset, add_missing. Returns: result
check_junctionsA
Check junction thinning consistency across related glyphs.
Measures how stems thin at arch/bowl junctions (n, m, b, d, p, q, etc.) by sweeping horizontal rays at multiple heights. Reports thinning % (100% = no thinning, 80% = stem thins to 80% at junction).
Checks consistency within groups (n≈m, b≈p, d≈q). Does NOT flag based on absolute values — thinning is design-specific. Only flags inconsistencies between related forms.
If master_id is omitted, analyzes ALL masters.
Does not modify glyph colors by default. When mark_glyphs=True, marks
inconsistencies red and passing glyphs green.
Args: glyph_names: Optional list of glyphs to check (default: n,h,m,u,a,b,d,p,q) master_id: Optional master ID (empty = all masters) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: master_id, glyph_names, mark_glyphs. Returns: result
check_language_supportA
Check the open font's language support with Shaperglot.
Exports a temporary binary from the active Glyphs font, then runs
Shaperglot against it. With no languages provided, returns the
languages whose score meets support_threshold. With languages,
returns per-language scores plus warnings/failures.
Args: languages: Optional language IDs or names, e.g. ["en_Latn", "Turkish", "Navajo"] instance_name: Optional Glyphs instance name to export. Empty = first exportable instance. export_format: "otf" (default) or "ttf" support_threshold: Minimum score to count a language as supported in report mode max_results: Max languages to show in report mode
Inputs: languages, max_results, export_format, instance_name, +1 more. Returns: result
check_overshootsA
Check overshoot values for round and pointed forms.
RECIPE: For a pre-export check, follow get_recipe("master_compatibility").
Round forms (O, o, C, S, etc.) should overshoot baseline and zone top by ~1-2% of zone height. Pointed forms (A, V, W) need MORE overshoot than rounds to appear optically aligned.
If no glyph_names provided, checks all known overshoot-sensitive glyphs in the font (O,C,D,G,Q,S,U,A,V,W,M,N,o,c,e,s,b,d,p,q,g,a,u,v,w,y + figures).
If master_id is omitted, analyzes ALL masters.
Does not modify glyph colors by default. When mark_glyphs=True, marks
missing/excessive overshoots red and passing glyphs green.
Args: glyph_names: Optional list of glyphs to check (default: all overshoot glyphs) master_id: Optional master ID (empty = all masters) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: master_id, glyph_names, mark_glyphs. Returns: result
check_punctuationA
Check punctuation consistency: mirrored pairs, width matches, and ratio checks.
Based on industry patterns across professional fonts. Checks:
Width matches (should be identical/similar):
Mirrored pairs: parenleft/parenright, bracketleft/bracketright, braceleft/braceright, guillemotleft/guillemotright (high severity — must match)
Related pairs: period/comma, colon/semicolon, quotedblleft/quotedblright (medium)
Width ratios (expected relationships):
endash wider than hyphen (traditionally 2x)
emdash wider than endash (traditionally 2x)
quoteright similar width to comma
exclam narrower than question
Skips any pairs where glyphs are missing. If master_id is omitted, analyzes ALL masters.
Does not modify glyph colors by default. When mark_glyphs=True, marks
mismatches red, width warnings yellow, and passing glyphs green.
Args: master_id: Optional master ID (empty = all masters) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: master_id, mark_glyphs. Returns: result
compare_colorA
Compare typographic color (ink density) across multiple glyphs.
RECIPE: For a full consistency audit, follow get_recipe("audit_consistency").
Finds glyphs that are visually too dark or too light compared to the group. Uses per-glyph expected density ratios from industry patterns — each glyph gets a verdict: pass, compensation, inconsistent, or unreliable.
Reference glyphs: n (lowercase), H (uppercase).
If master_id is omitted, analyzes ALL masters and returns per-master results. Glyph color in GlyphsApp = worst verdict across all masters.
Does not modify glyph colors by default. When mark_glyphs=True, marks:
Red (0) = inconsistent density
Orange (1) = unreliable measurement
Yellow (3) = optical compensation (expected)
Green (4) = pass
Args: glyph_names: Glyphs to compare master_id: Optional master ID (empty = all masters) mark_glyphs: Color glyphs by verdict only when explicitly requested
Examples: "Is the color consistent across uppercase?" -> compare_color(["H","I","M","N","O","B","D","E","F","K","L","P","R"])
Inputs: glyph_names (required), master_id, mark_glyphs. Returns: result
compare_proportionsA
Compare width proportions across glyphs within a font.
RECIPE: For a full consistency audit, follow get_recipe("audit_consistency").
Checks three things:
Related-form groups: b≈d≈p≈q (mirrored), h≈n≈u (arch), O≈Q, etc.
Width ordering: m>n>r, W>H>I — flags violations
Industry ranges from 18 professional fonts — flags outliers
References: n for lowercase, H for uppercase/figures.
If no glyph_names provided, checks all LC + UC + figures. If master_id is omitted, analyzes ALL masters.
Does not modify glyph colors by default. When mark_glyphs=True, marks
inconsistencies red, range warnings yellow, and passing glyphs green.
Args: glyph_names: Optional list of glyphs to check master_id: Optional master ID (empty = all masters) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: master_id, glyph_names, mark_glyphs. Returns: result
compare_stemsA
Compare stem thicknesses across multiple glyphs to find inconsistencies.
RECIPE: For a full font audit (stems + color + proportions + spacing), call get_recipe("audit_consistency") — it runs checks in dependency order.
Evaluates each glyph against industry stem patterns for optical compensation. Per-glyph verdicts:
pass: stem within expected tolerance (green in GlyphsApp)
compensation: known optical compensation like round stems, bowl mass (yellow) — only flagged if OUTSIDE the expected range
inconsistent: real issue, deviation exceeds industry norms (red)
unreliable: glyph shape can't be reliably measured (orange)
Each glyph is classified into one of 5 groups:
straight: pure stems (n,h,m,u,i,j,l,r / H,I,L,T,U,F,E,K,J)
round: pure round forms (o,c / O,C,Q)
mixed: stem + bowl (b,d,p,q,g,a,e,s / D,B,P,R,G)
diagonal: diagonal strokes (v,w,x,y,z,k / V,W,X,Y,Z,A,M,N)
optical/figure: special cases (t,f) and numbers (0-9)
Supports uppercase (ref: H), lowercase (ref: n), and figures (ref: H).
If master_id is omitted, analyzes ALL masters and returns per-master results. Glyph color = worst verdict across all masters. If master_id is provided, analyzes that single master only.
Does not modify glyph colors by default. When mark_glyphs=True, marks:
Red=inconsistent, Orange=unreliable, Yellow=compensation, Green=pass
Args: glyph_names: List of glyphs to compare master_id: Optional master ID (empty = all masters) mark_glyphs: Color glyphs by verdict only when explicitly requested
Inputs: glyph_names (required), master_id, mark_glyphs. Returns: result
create_glyphA
Create a new glyph in the open font.
Args: glyph_name: Name for the new glyph (e.g., "a.ss01", "uni0041") width: Advance width in font units unicode_value: Optional unicode value (e.g., "0061" for 'a') paths: Optional initial paths (same format as set_glyph_paths)
Inputs: glyph_name (required), paths, width, unicode_value.
create_recipeA
Create a new workflow recipe.
Recipes are markdown files with numbered steps. Follow this format:
Recipe: My Recipe Title
Description of what this recipe does.
Steps
1. First step title
tool_name— what it doesanother_tool— with parameters
2. Second step title
tool_name— instructions
Args: name: Snake_case identifier (e.g. 'fix_spacing_issues') content: Full markdown content of the recipe overwrite: If true, replace existing recipe with same name
Inputs: name (required), content (required), overwrite. Returns: result
delete_glyphA
Delete a glyph from the font. This cannot be undone via MCP.
Inputs: glyph_name (required).
delete_kerning_pairB
Delete a kerning pair.
Args: left: Left glyph name or group (@MMK_L_...) right: Right glyph name or group (@MMK_R_...) master_id: Optional master ID (uses first master if empty)
Inputs: left (required), right (required), master_id.
delete_recipeC
Delete a workflow recipe.
Args: name: Recipe name to delete (e.g. 'old_workflow')
Inputs: name (required). Returns: result
duplicate_glyphB
Duplicate a glyph (all layers and paths) under a new name.
Useful for creating alternates or backups before editing.
Inputs: new_name (required), glyph_name (required).
execute_in_glyphsA
Execute arbitrary Python code inside GlyphsApp.
The code runs with access to the Glyphs object and all GlyphsApp API. stdout is captured and returned. This is powerful but must be used carefully.
NOTE: This endpoint is disabled by default. The user must enable it in GlyphsApp preferences (com.nico.glyphs-mcp.allowExecute = True).
RECIPE: Do not improvise multi-step work here. To create a Glyphs plugin or script, follow get_recipe("create_glyphs_plugin") or get_recipe("create_glyphs_script"). For bulk layer cleanup, follow get_recipe("cleanup_dated_layers").
Example: execute_in_glyphs("print(Glyphs.font.familyName)")
Inputs: code (required).
export_fontA
Export the open font beside its saved source with glyphs-cli.
Creates export/YYYY-MM-DD_HH-MM-SS next to the open .glyphs or
.glyphspackage source. Static instances are exported as OTF, TTF,
WOFF, and WOFF2. Variable instances are exported as variable TTF files.
Previous export directories are never removed or overwritten.
The response always contains portable exportLog and colored
exportLogAnsi variants. The calling agent MUST show the ANSI variant
when its client supports terminal colors, otherwise the portable log,
after every export attempt including failures.
Args: save_before_export: Save pending changes before exporting. Defaults to False; if the document is edited, the tool asks for confirmation instead of silently saving it. timeout: Maximum seconds for each of the three export runs.
Inputs: timeout, save_before_export.
generate_box_drawingA
Generate Unicode box drawing and block element glyphs in the open font.
By default, processes U+2500–U+259F. Existing drawn glyphs are skipped
unless overwrite=True. The tool draws directly in the .glyphs source
across all masters, using auto-detected stem values when stroke params
are left at 0.
Args: glyph_names: Optional subset to generate. Accepts glyph names, single characters, or codepoints like "U+2500". overwrite: Replace existing drawings if True. width: Force advance width for generated glyphs. 0 = auto. stroke: Light/single stroke thickness. 0 = auto. heavy_stroke: Heavy stroke thickness. 0 = auto. double_gap: Gap between double-line strokes. 0 = auto. color: Optional glyph color label for newly created glyphs.
Inputs: color, width, stroke, overwrite, +3 more.
get_featuresB
Get OpenType feature code from the font.
Inputs: no named parameters.
get_font_infoA
Get information about the currently open font in GlyphsApp.
Returns font family name, units per em, glyph count, masters, axes, metrics (ascender, descender, x-height, cap height), and instances.
Inputs: no named parameters.
get_glyphA
Get complete data for a specific glyph including all paths and metrics.
Returns all layers with: paths (nodes with x,y coordinates, type, smooth), components, anchors, width, sidebearings.
Node types: "line" (straight on-curve), "curve" (smooth on-curve), "offcurve" (cubic bezier control point). Path direction: -1 = counter-clockwise (outer contours), 1 = clockwise (inner contours/counters).
Inputs: glyph_name (required).
get_glyph_svgC
Get a glyph rendered as SVG markup.
Returns SVG with viewBox matching the glyph's metrics. Useful for seeing the glyph shape as a vector image.
Inputs: glyph_name (required), master_id.
get_kerningA
Get kerning pairs for a specific master (or first master if not specified).
Args: master_id: Master ID. Empty = first master. left: Filter by left glyph/group name. Empty = all. limit: Max pairs to return. 0 = all. Use limit=50 for large fonts.
Inputs: left, limit, master_id.
get_mastersA
Get all font masters with their metrics and axis positions.
Returns master ID, name, ascender, descender, x-height, cap-height, and axis values for each master.
Inputs: no named parameters.
get_recipeA
Get a specific workflow recipe by name.
Returns the full recipe with step-by-step instructions, plus totalSteps count. Use get_recipe_step() to read and execute one step at a time.
Args: name: Recipe name (from list_recipes, e.g. 'scale_proportions')
Inputs: name (required). Returns: result
get_recipe_stepA
Get a single step from a recipe for sequential execution.
Returns step title, content, and a MANDATORY directive field. You MUST follow the directive — it tells you exactly what to do next.
The response includes:
step: current step number
totalSteps: total steps in recipe
title: step title
content: tools to call and instructions
directive: MANDATORY instruction for what to do after this step
IMPORTANT: Execute ALL tools listed in the step content before proceeding. Do NOT skip steps. Do NOT combine steps. Follow the directive field.
Args: name: Recipe name (e.g. 'audit_consistency') step: Step number (1-based)
Inputs: name (required), step (required). Returns: result
get_selectionA
Get the user's current selection in GlyphsApp's editor.
Returns the active glyph name, layer, and any selected paths or nodes. Returns null selection if no tab or layer is active.
Inputs: no named parameters.
get_spacing_stringsA
Get spacing test strings for visually evaluating a glyph's spacing.
RECIPE: For systematic spacing work, follow get_recipe("spacing_workflow").
Generates canonical test strings based on industry-standard methods:
Three-at-a-time (OH no Type Co): glyph sandwiched between n/o or H/O
Systematic pairs (Jamra): glyph paired with every letter in its case
Cross-case integration: glyph in mixed UC/LC context
Ruder test: hard vs easy word columns for overall color evaluation
Single-stem stress test: words like "millennial", "minimum" (for i, l, r, t)
Use these strings in GlyphsApp's Edit view to visually assess spacing quality.
Args: glyph_name: Name of the glyph to generate test strings for
Inputs: glyph_name (required). Returns: result
get_stem_targetsA
Get the designer's intended stem values from the Dimensions palette.
Returns stem snap values from custom parameters AND measured values from reference glyphs (H, O, n, o). These are the TARGET values that other glyphs should match.
Use this as baseline for compare_stems — to check if glyphs hit the designer's intended values, not just if they're consistent with each other.
Inputs: master_id.
list_glyphsA
List glyphs in the open font with basic metadata.
Returns glyph names, unicodes, category, subCategory. Does NOT include path data — use get_glyph() for full details.
Args: category: Filter by category — "Letter", "Number", "Punctuation", "Symbol", "Separator", "Mark", "Other". Empty = all. limit: Max glyphs to return. 0 = all.
Inputs: limit, category.
list_recipesA
List available workflow recipes.
Recipes are step-by-step guides for complex type design tasks. They tell you which tools to call, in what order, and what to check.
IMPORTANT: You MUST call this before starting any multi-step type design task (scaling, auditing, spacing, kerning, compatibility checks, etc.). If a recipe matches the task, call get_recipe(name) and follow its steps in order — they encode expert type design knowledge and dependency ordering.
Inputs: no named parameters. Returns: result
measure_colorA
Measure the ink density (typographic color) of a single glyph.
Ink density = ratio of filled area to total bounding box. Normalizes by the appropriate zone height (x-height for lowercase, cap-height for uppercase) so values are comparable across glyphs.
Returns a value between 0.0 and 1.0 (typically 0.15-0.50 for text fonts).
Higher = darker/heavier glyph. Lower = lighter.
Inputs: glyph_name (required), master_id.
measure_stemsA
Measure stem thicknesses in a glyph using ray-casting.
Casts horizontal and vertical rays through the glyph outlines to measure vertical stems (like the main strokes of n, m, h) and horizontal stems (like crossbars of e, A, H).
Returns: - Dominant (most common) vertical and horizontal stem values - Min/max range showing consistency within the glyph - Detailed per-ray measurements with positions
Use this BEFORE and AFTER applying RMX filters to verify the effect.
Examples: "How thick are the stems of R?" -> measure_stems("R") "What's the crossbar thickness of H?" -> measure_stems("H")
Inputs: glyph_name (required), samples, master_id.
rename_glyphA
Rename a glyph. Fails if new_name already exists.
Inputs: new_name (required), glyph_name (required).
review_productionA
Run a comprehensive production readiness review on the open font.
Checks 44 items across critical/warning/info severities:
Critical (must fix): family name, .notdef, space glyph, duplicate unicodes/names, master compatibility, vertical metrics, open paths, valid glyph names, missing components, alignment zones.
Warning (should fix): metadata, version, essential glyphs, .notdef outlines, zone overshoots, Use Typo Metrics, typo/hhea match, win metrics coverage, kerning groups, cross-master kerning, critical kern pairs, OT features, weight classes, style linking, nbspace width, zero-width letters, PANOSE, stems defined, short segments, near-miss alignments, presentation forms decomposition (fi/fl/ff), German sharp S, Dutch IJ localization, smallcap completeness, salt feature, languagesystems prefix, Remove Overlap.
Info: glyph count by category, font metrics, fsType, VF readiness, unreachable glyphs.
Returns a formatted markdown report with pass/fail status and details for failures.
Inputs: no named parameters. Returns: result
rmx_batchA
Apply an RMX filter to multiple glyphs at once.
RECIPE: For multi-glyph scaling, call get_recipe("scale_proportions") first.
Args: filter_name: One of "harmonize", "tune", "scale", "monospace" glyph_names: List of glyph names to process params: Filter-specific parameters (same as individual tool params) master_id: Optional master ID
Examples: "Harmonize all lowercase" → rmx_batch("harmonize", ["a","b",...,"z"], params={"mode": "harmonize"})
"Make all caps 20% wider" →
rmx_batch("scale", ["A","B",...,"Z"], params={"width": 120})
Inputs: filter_name (required), glyph_names (required), params, master_id.
rmx_harmonizeB
Optimize bezier curves on a glyph using RMX Harmonizer.
Modes:
"harmonize": Full curve optimization (recommended default)
"dekink": Only fix kinks at smooth connections
"extract handles": Reset handles to default positions
"supersmooth diagonals": Extra smoothing on diagonal segments
"supersmooth all": Maximum smoothing everywhere
Use after drawing or modifying paths to ensure clean curves. Works on any glyph — no multi-master requirement.
Inputs: glyph_name (required), mode, master_id.
rmx_monospaceA
Adjust a glyph to a fixed width using RMX Monospacer.
Intelligently distributes width change between outline scaling and spacing adjustment to maintain visual quality.
Args: glyph_name: Name of the glyph mono_width: Target advance width (0 = keep current width) keep_stroke: How much to preserve stroke weight, 0-100% use_spacing: How much width change goes to spacing vs outline, 0-100% master_id: Optional master ID
Example: "Make all uppercase letters 600 units wide" → rmx_batch("monospace", ["A","B",...,"Z"], params={"monoWidth": 600})
Requires 2+ masters.
Inputs: glyph_name (required), master_id, mono_width, keep_stroke, +1 more.
rmx_scaleA
Scale a glyph by percentage in width and/or height.
USE THIS for percentage-based scaling requests like "make 30% wider".
RECIPE: For a multi-glyph scaling task, call get_recipe("scale_proportions") first and follow it step by step.
Uses RMX Scaler with stroke-weight compensation via master interpolation. Native affine fallback is opt-in because it cannot preserve stem weight. Response includes a "method" field: "rmx" = RMX Scaler was used, "native_transform" = affine transform fallback.
Args: glyph_name: Name of the glyph to scale width: Width scale as percentage. 100 = no change, 130 = 30% wider, 70 = 30% narrower height: Height scale as percentage. 100 = no change weight: Stroke weight delta (RMX only, ignored in native fallback) adjust_space: Sidebearing adjustment delta vertical_shift: Vertical position offset master_id: Optional active master ID. RMX still processes all masters. allow_fallback: Permit native affine scaling if RMX fails. Default False.
RMX parameters accept either one value for every master or a list in
font master order for independent per-master control.Examples: "Make R 30% wider" → rmx_scale("R", width=130) "Make R 20% narrower" → rmx_scale("R", width=80) "Scale R to 90% height, keep width" → rmx_scale("R", height=90)
Inputs: glyph_name (required), width, height, weight, +4 more.
rmx_tuneA
Adjust a glyph by delegating directly to the installed RMX Tuner.
This tool calls the real RMXTuner instance loaded in GlyphsApp
(Glyphs.filters) and passes the values to Tuner's own controls. It does
NOT approximate Tuner with native interpolation.
IMPORTANT: Width/height/weight/slant are RMX Tuner deltas, not percentages. Typical useful range is roughly -100 to +100, depending on the font/RMX setup. Blend is normalized: 0 = no blend-in layer, 1 = full blend-in layer. Values above 1 extrapolate and can produce extreme outlines.
Args: glyph_name: Name of the glyph to tune. weight: RMX Tuner Weight value (+ = heavier, - = lighter). width: RMX Tuner Width value (+ = wider, - = narrower). height: RMX Tuner Height value (+ = taller, - = shorter). slant: RMX Tuner Slant value. blend: RMX Tuner Blend value. Use 0.0–1.0 for interpolation. fixed_width: Keep advance width fixed via RMX Tuner's checkbox. master_id: Optional master/layer ID. Empty = first master. all_masters: If True, pass all master layers of the glyph to RMX Tuner.
Example: rmx_tune("R.001", height=88, weight=88)
Inputs: glyph_name (required), blend, slant, width, +5 more.
set_feature_codeB
Create or update an OpenType feature.
Args: feature_name: Feature tag e.g. "liga", "kern", "ss01" code: OpenType feature code (AFDKO syntax) active: Whether the feature is enabled (default True)
Inputs: code (required), feature_name (required), active.
set_glyph_colorA
Set the color label of a glyph in the font view.
Color index (0–11): 0=red, 1=orange, 2=brown, 3=yellow, 4=light green, 5=dark green, 6=teal, 7=blue, 8=purple, 9=pink, 10=light gray, 11=charcoal
Use None/no color by setting any value outside this range.
Inputs: color (required), glyph_name (required).
set_glyph_pathsA
Replace ALL paths on a glyph's layer in GlyphsApp.
This OVERWRITES existing paths. The glyph must already exist.
Each path is: {"closed": true, "nodes": [{"x": 100, "y": 0, "type": "line", "smooth": false}, ...]}
Node types:
"line": on-curve point with straight connection to previous
"curve": on-curve point with smooth cubic bezier connection
"offcurve": cubic bezier control handle (always in pairs before a "curve")
Curve segments: offcurve, offcurve, curve (3 nodes per segment)
Path direction: outer contours counter-clockwise, counters/holes clockwise. correctPathDirection() is called automatically after setting paths.
If master_id is empty, uses the first master.
Inputs: paths (required), glyph_name (required), master_id.
set_glyph_unicodeB
Set the unicode value of a glyph.
Args: glyph_name: Name of the glyph to update unicode_value: Hex unicode string e.g. "0061" for 'a'. Pass "" to clear.
Inputs: glyph_name (required), unicode_value (required).
set_glyph_widthC
Set the advance width of a glyph.
Inputs: width (required), glyph_name (required), master_id.
set_kerning_pairA
Set a kerning pair value between two glyphs.
Args: left: Left glyph name or group (@MMK_L_...) right: Right glyph name or group (@MMK_R_...) value: Kerning value (negative = tighter) master_id: Optional master ID (uses first master if empty)
Inputs: left (required), right (required), value (required), master_id.
smart_scaleA
Scale glyphs with automatic stem weight compensation.
RECIPE: For multi-glyph scaling tasks, call get_recipe("scale_proportions") first — it includes pre/post stem verification and visual comparison steps.
Unlike simple affine transforms which distort stem weights, this tool:
Measures reference stems (H for UC, n for LC) before scaling
Applies the width/height transform
Uses GlyphsFilterOffsetCurve to compensate stem thickness changes
Reports before/after stem measurements for verification
All values are scale factors where 1.0 = no change.
Args: width: Horizontal scale factor. 0.97 = 3% narrower, 1.1 = 10% wider height: Vertical scale factor. 1.15 = 15% taller, 0.9 = 10% shorter weight: Target stem weight factor. 1.0 = maintain original stem thickness after scaling (compensate). 0.9 = make stems 10% thinner. 1.1 = 10% thicker. proportional: If true, height follows width (uniform scale with compensation) glyph_names: List of glyph names to process (empty = all exporting glyphs) master_id: Process only this master (empty = all masters) backup: Create backup layers before modifying (default true)
Examples: Condense 3% keeping weight: smart_scale(width=0.97, weight=1.0) Increase x-height 15% keeping weight: smart_scale(height=1.15, weight=1.0) Scale uniformly 90% with weight compensation: smart_scale(width=0.9, proportional=True) Make all stems 10% thicker: smart_scale(weight=1.1) Condense UC only: smart_scale(width=0.95, glyph_names=["A","B","C",...])
Inputs: width, backup, height, weight, +3 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 55 tools
Most tools target a distinct resource, but there is real overlap among scaling/transformation tools such as rmx_scale, smart_scale, rmx_tune, and rmx_batch when used with the "scale" filter. The detailed descriptions help, but an agent could still easily pick the wrong entry point for a similar-looking operation.
The vast majority of tools follow a readable snake_case verb_noun pattern like get_glyph, set_kerning_pair, and analyze_spacing. The rmx_*, smart_*, and auto_* prefixed tools break the pure pattern slightly, but the conventions are internally consistent enough to stay predictable.
With 55 tools, this clearly exceeds the 50+ threshold for an extreme tool count. Many audit, comparison, transformation, and recipe tools could be consolidated, and the recipe meta-tools add a large amount of surface area that could be folded into a smaller set.
The toolset provides strong coverage across glyph CRUD, paths, widths, unicode, colors, kerning, features, masters, exports, and a wide range of design audits. Minor gaps exist, such as no direct component/anchor editing, no sidebearing setter, and no feature deletion, but most workflows can still be completed.
Maintenance
Related MCP Connectors
Talk to your public-facing AI from any MCP client — Claude, ChatGPT, Cursor, Cline, Windsurf.
Real-time chat for AI agents. Claude Code, Cursor, Cline and Codex join channels over MCP.
Real-time chat hub for AI agents — Claude Code, Cursor, Cline, Codex over MCP or REST.
One memory for all your AI tools: Claude, ChatGPT, Cursor and any MCP client share the same notes.
Related MCP Servers
- AlicenseAqualityDmaintenanceWrite-side MCP server for Figma — build, edit, and search Figma designs from Claude Code, Cursor, Cline, or any MCP client. Complements Figma's official read-only MCP with 41 tools for tree creation, variables, components, and visual verification.413MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI assistants to interact with FontLab for font design and manipulation, including querying font metadata, creating/modifying glyphs, applying transformations, and exporting fonts.8MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI tools to create and edit Font Proof documents for type designers, supporting PDF font proofs and live-reload from Glyphs.app.27 npmMIT
- AlicenseNot gradedqualityAmaintenanceA Model Context Protocol server for Glyphs that exposes font‑specific tools to AI/LLM agents.60Apache 2.0