TokenMoulds

A corporate template system for Microsoft Office and LibreOffice. Point it at a brand, get publication-quality Word, PowerPoint, Excel, and LibreOffice templates — with embedded fonts, baseline grids, OOXML compliance validation, and long-form document support for whitepapers, annual reports, manuals, and bound books.

Install

pip install tokenmoulds

Quick Start

Generate templates from brand inputs:

tokenmoulds --org-id="acme" \
  --font-pair="inter-roboto" \
  --primary-color="#2563EB" \
  --secondary-color="#DC2626" \
  --locale="GB" \
  --brand-tone=0.3 \
  --generate-templates --generate-odf

Or from a DTCG token file:

tokenmoulds --org-id="acme" \
  --tokens-file="design.tokens.json" \
  --generate-templates

What It Does

~30 brand inputs (colors, fonts, tone) produce 1,300+ derived design tokens across the current Office, ODF, and Google Workspace outputs. Every template ships with:

• WCAG AAA accessible contrast ratios
• Embedded fonts with ODTTF obfuscation (Word) and EOT with GSUB (PowerPoint)
• OpenType typography: contextual figures, size-dependent tracking, ligatures
• Baseline grid-snapped typography via modular type scales
• ISO 29500 (OOXML) and ODF schema compliance
• Zero macros — all styling baked into document structure

Long-Form Documents

Corporate whitepapers, annual reports, manuals, and bound books need more than a single page layout. They need sections — front matter with roman-numeral page numbers, a body with running chapter titles and decimal numbering, an appendix that continues the count. Each section with its own headers, footers, and page geometry.

TokenMoulds generates long-form structure from a small token vocabulary and emits it consistently across both Word (.dotx) and LibreOffice Writer (.ott) from one shared IR. Features:

• Per-section page geometry — A4, A5, B5, Letter, portrait or landscape, margins as formulas against the baseline grid
• Mirror margins — inner/outer semantics for bound documents, facing pages automatically flip
• Running headers — STYLEREF:Heading1 fields that update as the reader moves through chapters (Word) / <text:chapter> fields (Writer)
• Per-section page numbering — lowerRoman for front matter, decimal restart for the body, continuing counts for appendices
• Odd/even headers — book-style layouts with different left-hand and right-hand page content
• First-page different — cover pages, chapter openers
• Multi-column sections — indexes, newsletters
• Cross-format parity — one IR → structurally identical .dotx and .ott, locked down by a golden master test suite

See Long-Form Documents guide for a walkthrough with a three-section whitepaper example.

Output Formats

MCP Server

For AI-assisted document creation via Claude, Cursor, or other MCP clients:

tokenmoulds mcp-server

48 tools for corporate GitHub linkage, artifact planning, validation, compilation, quality gating, artifact review bundles, GitHub PR publishing, Google Apps Script generation, direct Apps Script API deployment, Google Workspace import/export roundtrips, python-pptx intent translation, brand extraction, template generation, document creation, validation, and cache management. See the MCP tool catalog.

For a direct repo-native Google Apps Script handoff, generate the Apps Script bundle from a stored plan:

tokenmoulds google-suite --artifact-id quarterly-update --output-dir ./generated

For Google Workspace import/export evidence, roundtrip a local artifact through Drive conversion:

tokenmoulds google-roundtrip generated/acme/deck.pptx --folder-id "$TOKENMOULDS_GSUITE_FOLDER_ID"

Python API

from tokenmoulds.pipeline import BuildConfig, build

result = build(BuildConfig(
    org_id="acme",
    output_formats=["potx", "dotx", "xltx"],
    brand_inputs={
        "font_pair": {"sans": "Inter", "serif": "Roboto Slab"},
        "base_colors": {"primary": "#2563EB", "secondary": "#DC2626"},
        "locale": "US",
        "brand_tone": 0.5,
    },
))

for fmt, data in result.artifacts.items():
    open(f"acme.{fmt}", "wb").write(data)

Architecture

TokenMoulds uses a single canonical build pipeline:

Design Tokens → DesignResolutionEngine → DocumentIR → Emitters → Validated Packages

• DTCG pipeline: W3C Design Tokens Community Group format with 5-layer resolution (core/fork/org/group/personal)
• Design engine: Modular type scale, baseline grid, OpenType features, tone/density adaptation
• Document IR: Format-agnostic intermediate representation
• 11 emitters: Each produces a complete, valid package
• Validation: In-process ISO 29500 schema checking via openxml-audit

See ADR 027 for the full convergence story.

Development

git clone https://github.com/BramAlkema/TokenMoulds.git
cd TokenMoulds
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
python -m pytest tests/ -v --tb=short   # 2,200+ tests

Contributing

See CONTRIBUTING.md for development guidelines, code conventions, and how to submit changes.

Documentation

• Getting Started
• CLI Reference
• MCP Tool Catalog
• Architecture Decisions

License

MIT