Typst MCP Server is an MCP (Model Context Protocol) implementation that helps AI models interact with Typst, a markup-based typesetting system. It exposes tools for LaTeX→Typst conversion, Typst syntax validation, docs lookup and Typst → PNG rendering.
https://github.com/johannesbrandenburger/typst-mcpYour AI coding assistant knows LaTeX inside and out, but ask it to generate Typst and you'll get syntax errors and frustrated iterations. This MCP server bridges that gap by giving AI models the tools they need to work confidently with Typst.
Here's the reality: Typst is cleaner and more modern than LaTeX, but AI models were trained on decades of LaTeX documents. When you ask Claude or Cursor to help with Typst documents, you're fighting against their training data.
This MCP server flips that dynamic. Instead of forcing AI to learn Typst syntax from scratch, it lets them write what they know (LaTeX) and converts it automatically. The result? Accurate Typst code without the back-and-forth debugging.
LaTeX → Typst Conversion: AI writes LaTeX (which it's excellent at), server converts to clean Typst markup. No more syntax errors from AI trying to remember Typst's function syntax.
Syntax Validation: Before your AI assistant sends you Typst code, it validates the syntax. Catch errors in the AI conversation, not in your compiler.
Visual Feedback: For complex layouts and mathematical expressions, the server renders Typst to PNG so multimodal AI can actually see what it created and iterate on the visual output.
Live Documentation Access: AI can query specific Typst documentation chapters on-demand, getting the exact syntax reference it needs for your current task.
Academic Paper Collaboration: You're writing a paper with complex mathematical notation. Ask your AI assistant to generate the equations in LaTeX (which it nails), then convert to Typst with proper syntax validation before integration.
Technical Documentation: Building diagrams and structured layouts in Typst. The AI can render previews and iterate on the visual output until the layout is exactly what you need.
Learning Typst: Instead of manually translating LaTeX examples to Typst, let your AI assistant handle the conversion while you focus on understanding the differences.
Works with Claude Desktop, VS Code agent mode, and Cursor right now. The setup is straightforward - clone, install dependencies (Python, Typst CLI, Pandoc), and run the server. Your AI assistant immediately gains access to all conversion and validation tools.
The server runs locally, so your documents stay private while giving AI the Typst capabilities it needs to be genuinely helpful rather than a source of syntax frustration.
You chose Typst because it's better than LaTeX for modern document workflows. But training AI assistants on every typesetting system isn't practical. This server lets you keep the benefits of Typst while leveraging AI's existing LaTeX expertise.
The validation and rendering feedback loop means fewer compilation errors and more time focused on content rather than syntax debugging. Your AI assistant becomes a Typst expert without you having to become one first.