Typst MCP Server
Enables AI models to interact with the Typst typesetting system, including LaTeX-to-Typst conversion, syntax validation, and rendering Typst code to images.
README
Typst MCP Server
Typst MCP Server is an MCP (Model Context Protocol) implementation that helps AI models interact with Typst, a markup-based typesetting system. The server provides tools for converting between LaTeX and Typst, validating Typst syntax, and generating images from Typst code.
Available Tools
⚠️ Currently all the functionality is implemented as
tools, because Cursor and VS Code are not able to handle the other primitives yet.
The server provides the following tools:
-
list_docs_chapters(): Lists all chapters in the Typst documentation.- Lets the LLM get an overview of the documentation and select a chapter to read.
- The LLM should select the relevant chapter to read based on the task at hand.
-
get_docs_chapter(route): Retrieves a specific chapter from the Typst documentation.- Based on the chapter selected by the LLM, this tool retrieves the content of the chapter.
- Also available as
get_docs_chapters(routes: list)for retrieving multiple chapters at once.
-
latex_snippet_to_typst(latex_snippet): Converts LaTeX code to Typst using Pandoc.- LLMs are better at writing LaTeX than Typst, so this tool helps convert LaTeX code to Typst.
- Also available as
latex_snippets_to_typst(latex_snippets: list)for converting multiple LaTeX snippets at once.
-
check_if_snippet_is_valid_typst_syntax(typst_snippet): Validates Typst code.- Before sending Typst code to the user, the LLM should check if the code is valid.
- Also available as
check_if_snippets_are_valid_typst_syntax(typst_snippets: list)for validating multiple Typst snippets at once.
-
typst_to_image(typst_snippet): Renders Typst code to a PNG image.- Before sending complex Typst illustrations to the user, the LLM should render the code to an image and check if it looks correct.
- Only relevant for multi modal models.
Getting Started
-
Clone this repository
git clone https://github.com/johannesbrandenburger/typst-mcp.git
-
Clone the typst repository
git clone https://github.com/typst/typst.git
-
Run the docs generation in the typst repository
cargo run --package typst-docs -- --assets-dir ../typst-mcp/typst-docs --out-file ../typst-mcp/typst-docs/main.json- Make sure to adjust the path to your local clone of the typst-mcp repository
- This will generate the
main.jsonand the assets in thetypst-docsfolder
-
Install required dependencies:
uv sync(install uv if not already installed) -
Install Typst
Running the Server
Local Installation
Execute the server script:
python server.py
Or install it in Claude Desktop with MCP:
mcp install server.py
Or use the new agent mode in VS Code:
Agent mode: available to all users and supports MCP
Docker
For Docker installation and usage, see DOCKER.md for detailed build information and instructions.
Platform Configuration
This MCP server can be integrated with various AI coding platforms. Below are configuration instructions for popular platforms:
Claude Desktop
Add the following to your Claude Desktop configuration file (claude_desktop_config.json):
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"typst": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"ghcr.io/johannesbrandenburger/typst-mcp:latest"
]
}
}
}
Or for local installation:
{
"mcpServers": {
"typst": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"ghcr.io/johannesbrandenburger/typst-mcp:latest"
]
}
}
}
Cursor
Create .cursor/mcp.json in your project root for project-specific configuration:
{
"mcpServers": {
"typst": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"ghcr.io/johannesbrandenburger/typst-mcp:latest"
]
}
}
}
Or create ~/.cursor/mcp.json for global configuration. For local installation:
{
"mcpServers": {
"typst": {
"command": "python",
"args": ["/path/to/typst-mcp/server.py"]
}
}
}
Roo Code
Create .roo/mcp.json in your project root for project-specific configuration:
{
"mcpServers": {
"typst": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"ghcr.io/johannesbrandenburger/typst-mcp:latest"
]
}
}
}
Or edit global settings via Roo Code MCP settings view. For local installation:
{
"mcpServers": {
"typst": {
"command": "python",
"args": ["/path/to/typst-mcp/server.py"]
}
}
}
OpenCode
Create or edit the opencode.json file in your project root:
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"typst": {
"type": "local",
"command": ["docker", "run", "--rm", "-i",
"ghcr.io/johannesbrandenburger/typst-mcp:latest"
],
"enabled": true
}
}
}
For local installation:
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"typst": {
"type": "local",
"command": ["python", "/path/to/typst-mcp/server.py"],
"enabled": true
}
}
}
Claude Code
Configure MCP servers in your Claude Code settings file (~/.claude/settings.json):
{
"mcpServers": {
"typst": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"ghcr.io/johannesbrandenburger/typst-mcp:latest"
]
}
}
}
Or for local installation:
{
"mcpServers": {
"typst": {
"command": "python",
"args": ["/path/to/typst-mcp/server.py"]
}
}
}
VS Code with Agent Mode
VS Code's Agent Mode has native MCP support (no extension required):
- Enable Agent Mode in VS Code settings (
chat.agent.enabled) - Create
.vscode/mcp.jsonin your workspace:
{
"servers": {
"typst": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"ghcr.io/johannesbrandenburger/typst-mcp:latest"
]
}
}
}
For global configuration, add to your user profile via MCP: Open User Configuration command.
JSON Schema of the Typst Documentation
⚠️ The schema of the typst documentation is not stable and may change at any time. The schema is generated from the typst source code and is not guaranteed to be complete or correct. If the schema changes, this repository will need to be updated accordingly, so that the docs functionality works again.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。