ipe-mcp
Enables AI agents to create, edit, validate, and render editable Ipe documents and presentations through semantic operations instead of raw XML.
README
ipe-mcp
ipe-mcp is a local-first, host-agnostic MCP server in development for creating,
editing, validating, and rendering editable Ipe
documents and presentations through AI agents.
The project combines a versioned semantic document model with deterministic Ipe
XML and native Ipe validation. Its baseline is Ipe 7.2.30 and XML format
70218. The initial target environment is Ubuntu 26.04 on WSL.
[!IMPORTANT] This repository is pre-MVP. Milestones M0–M5 are complete, covering the compatibility contracts, conformance lab, semantic IR, transactional persistence, layout, object authoring, and slide composition. The stdio MCP server is planned for M8, so the repository is not yet an installable MCP integration.
Goals
- Give agents semantic operations for pages, layers, views, objects, styles, layout, reveals, and export without requiring them to manipulate raw XML.
- Preserve editability and supported Ipe content across parse, mutation, save, native reload, and recovery.
- Keep layers, drawing order, and presentation views as separate concepts.
- Make every mutation revision-safe, atomic, validated, and recoverable.
- Support Codex and other MCP hosts through host-neutral contracts.
- Produce
.ipe, PDF, and raster preview artifacts with explicit validation and compatibility diagnostics.
Current Status
| Milestone | Scope | Status |
|---|---|---|
| M0 | Contracts, ADRs, and compatibility baseline | Complete |
| M1 | Ipe conformance laboratory | Complete |
| M2 | Semantic IR, XML, and transactional persistence | Complete |
| M3 | Coordinates and layout | Complete |
| M4 | Objects, geometry, text, assets, and styles | Complete |
| M5 | Pages, layers, views, and slide composition | Complete |
| M6 | Native validation, rendering, and export | Planned |
| M7 | Reveal, motion, scrolling, and viewer matrix | Planned |
| M8 | MCP stdio server and host integration | Planned |
| M9 | Hardening and MVP release candidate | Planned |
| M10 | Post-MVP extensions and distribution | Future |
See the roadmap for the complete scope, gates, risks, and design decisions.
Architecture
The implementation is organized around a few strict boundaries:
Agent or MCP host
|
v
Semantic operations and layout
|
v
Versioned document IR
|
+--> transactional session and sidecar
|
v
Deterministic Ipe XML codec
|
v
Ipe 7.2.30 native validation, rendering, and export
The normal API will expose typed document operations rather than arbitrary XML. Native Ipe tools remain the authority for behaviors that cannot be validated structurally, including style resolution, LaTeX, rendering, and export.
Key source areas:
src/domain: semantic IR, schemas, validation, and stable identities;src/ipe/xml: deterministic parser, projector, and serializer;src/persistence: bounded reads, atomic writes, sessions, snapshots, and sidecars;src/layout: coordinate spaces, matrices, anchors, constraints, layout, and connectors;fixtures/conformance: documents and golden data used against native Ipe.
Requirements
- Node.js 20 or later;
- npm;
- Ipe 7.2.30 for the supported full-validation path;
- pdfLaTeX for text compilation and native rendering workflows.
The verified environment uses Ubuntu's ipe 7.2.30-1build2 package on Ubuntu
26.04 WSL. Follow SETUP-WSL.md for installation and native-tool
verification.
Development Setup
git clone https://github.com/Mik1810/ipe-mcp.git
cd ipe-mcp
npm ci
npm run build
npm test
The package is currently private and has no published executable or MCP server entry point.
Verification
Run the milestone gates from the repository root:
bash scripts/check-m0.sh
bash scripts/check-m1.sh
bash scripts/check-m2.sh
bash scripts/check-m3.sh
The gates build on one another and cover structural checks, native Ipe round-trips, semantic fixed points, persistence, numerical behavior, and layout fixtures. M1 also supports an optional source-build lane:
IPE_M1_SOURCE_BIN_DIR=/path/to/ipe/build/bin bash scripts/check-m1.sh
Without that variable, the verified Ubuntu package is used and the optional source lane is reported as skipped.
Compatibility Modes
The design distinguishes three explicit modes:
- structural-only: parse, inspect, and generate with structural diagnostics, without claiming native verification;
- full 7.2.30: the supported release path with native validation;
- nightly 7.3.x: experimental compatibility, never used to rewrite a stable document without consent.
See docs/compatibility-modes.md for the precise capability and failure matrix.
Documentation
- ROADMAP.md: architecture, milestones, gates, and future work;
- docs/adr: accepted architecture decisions;
- docs/conformance-m1.md: native conformance lab;
- docs/core-m2.md: IR, XML, identity, and persistence;
- docs/core-m3.md: coordinates and layout;
- report-source.md: source dossier and traceability;
- ORCHESTRATOR_PROMPT.md: milestone execution and review protocol.
Project Principles
- Never mutate the original document before an explicit save.
- Treat every top-level object layer and every serialization-sensitive default explicitly.
- Validate semantically and with native Ipe; a file merely opening is not a sufficient gate.
- Preserve unknown supported content whenever possible and report any degraded behavior.
- Keep protocol contracts independent of Codex-specific prompts, skills, or UI directives.
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.