idml-mcp
Enables AI agents to parse, summarize, and safely edit Adobe InDesign IDML files without requiring InDesign.
README
idmlkit
The AI-native toolkit for Adobe InDesign files (IDML) — parse, summarize, and safely write back. No InDesign required.
AI tools have transformed how we write code and build web pages, but print and layout designers have largely been left out: the InDesign ecosystem has no AI-ready open interface. Existing InDesign MCP servers all remote-control a running InDesign instance — which means a license, a desktop, and an open app for every operation.
idmlkit takes the other road: it works directly on .idml files (InDesign's open, XML-based interchange format). That means it runs anywhere Node.js runs — servers, CI pipelines, Linux, a designer's laptop — with nothing from Adobe installed.
JSON data in ──► idmlkit ──► perfectly laid-out IDML out ──► open in InDesign / print
Design philosophy: the AI never touches the XML
LLMs are great at understanding a document and deciding what should change, and terrible at hand-editing 5,000 lines of layout XML without breaking it. So idmlkit splits the work:
- Deterministic code parses the package and produces a compact, LLM-friendly summary (pages, frames, stories as indexed text runs, image links, styles) — a few hundred tokens instead of megabytes of XML.
- Write-back is surgical: replacing a text run or relinking an image rewrites only that one element via targeted string surgery. Everything else — geometry, styles, the XML formatting itself — is preserved byte-for-byte (enforced by tests). The AI physically cannot corrupt your layout.
Validated against real InDesign 2025 exports: on Adobe's official Flex Layout sample (14.5 MB, with a single 20 MB spread), summarize runs in under half a second and a text edit changes exactly one zip entry — every other entry verified byte-identical by CRC.
Packages
| Package | What it is |
|---|---|
idmlkit |
Core library + CLI: open, summarize, safe text/image write-back |
idml-mcp |
MCP server exposing those abilities to Claude and any MCP-capable agent |
Quick start
npm install
npm run build
node scripts/make-sample-idml.mjs # generates examples/sample.idml
CLI
$ node packages/idmlkit/dist/cli.js summary examples/sample.idml
Pages: 1 Spreads: 1
Story u123 (11 words, 2 runs): Spring Catalog 2026 Fresh arrivals for the new season…
Image link u202: file:/images/hero.jpg
Paragraph styles: $ID/NormalParagraphStyle, Headline, Body
$ node packages/idmlkit/dist/cli.js set-text examples/sample.idml u123 0 "Summer Sale 2026" -o out.idml
$ node packages/idmlkit/dist/cli.js relink out.idml u202 "file:/images/summer.jpg"
MCP server (Claude Code / Claude Desktop)
{
"mcpServers": {
"idml": {
"command": "node",
"args": ["<repo>/packages/idml-mcp/dist/server.js"]
}
}
}
Then ask Claude things like "Summarize catalog.idml and replace every 2025 date with 2026" — it will use these tools:
| Tool | Purpose |
|---|---|
idml_summarize |
Structured summary of pages, frames, stories, images, styles |
idml_get_story |
One story's plain text + indexed runs with styles |
idml_set_text |
Replace one run's text (layout untouched, byte-for-byte) |
idml_relink_image |
Point an image link at a new file (geometry untouched) |
Library
import { IdmlDocument, summarizeDocument, setStoryRunText } from "idmlkit";
const doc = await IdmlDocument.open("catalog.idml");
console.log(summarizeDocument(doc)); // LLM-friendly JSON
setStoryRunText(doc, "u123", 0, "New headline");
await doc.save("catalog-updated.idml");
Use cases
- Database publishing: fill a designed template with product data — thousands of catalog pages, price lists, or menus without opening InDesign once.
- Localization: extract every text run with its styles, translate, write back safely.
- Agent workflows: let Claude read a layout, reason about its content, and apply text/image updates as part of a larger pipeline.
Roadmap
- [x] Read IDML packages, LLM-friendly summaries
- [x] Safe text run replacement and image relinking
- [x] MCP server (stdio)
- [ ] Page → frame mapping with geometry (which frame is where)
- [ ] Template fill: apply a JSON data object to a tagged template in one call
- [ ] Story-level structured diff (what changed between two IDML files)
- [ ] Overflow heuristics (warn when new text likely won't fit its frame)
- [ ] Real-world IDML corpus tests (CC 2018–2026 exports)
Development
npm install
npm run build # tsc project references
npm test # vitest — includes byte-for-byte preservation tests
License
MIT. Not affiliated with or endorsed by Adobe. "Adobe" and "InDesign" are trademarks of Adobe Inc., used here only to describe compatibility.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.