Overleaf MCP
Enables AI assistants to read, edit, and compile Overleaf LaTeX projects via the Overleaf web API using a browser session cookie.
README
Overleaf MCP
An MCP (Model Context Protocol) server that lets AI assistants like Claude read, edit, and compile your Overleaf projects.
MCP spec compatibility: This server is built on FastMCP 3.x, which targets the classic handshake-based MCP spec (2025-11-25 and earlier) — it does not yet implement the stateless 2026-07-28 spec ("MCP 2.0"). Current clients (Claude Code, Claude Desktop, etc.) still support handshake-era servers, so it works fine today. It will move to the stateless spec once FastMCP 4 is released.
Works with free Overleaf accounts — no premium Git integration required. It authenticates with your browser session cookie and talks to Overleaf's web API via pyoverleaf.
Tools
| Tool | Description |
|---|---|
list_projects |
List all projects on your Overleaf account |
list_files |
List files in a project (supports glob filters like *.tex) |
read_file |
Read any file from a project |
write_file |
Create or overwrite a file (preserves Overleaf history; creates folders as needed) |
compile_project |
Compile on Overleaf's servers; returns the log tail on failure |
get_compile_log |
Fetch the full LaTeX log from the latest compile |
download_pdf |
Download the compiled PDF to a local path |
Projects can be referenced by id, exact name, or a unique name substring — e.g. read_file("thesis", "main.tex").
Setup
1. Clone and install
git clone https://github.com/SudheerNaraharisetty/overleaf-mcp.git
cd overleaf-mcp
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt
2. Get your Overleaf session cookie
- Log in at www.overleaf.com
- Open browser DevTools (F12) → Application (Chrome) or Storage (Firefox) → Cookies →
https://www.overleaf.com - Copy the value of the
overleaf_session2cookie
3. Configure the cookie
Create a .env file next to server.py:
echo 'OVERLEAF_SESSION_COOKIE=<paste cookie value here>' > .env
Note: The cookie expires when your Overleaf session does. If the server starts returning auth errors, grab a fresh cookie and update
.env.
4. Add to your MCP client
Claude Code:
claude mcp add overleaf -- /path/to/overleaf-mcp/.venv/bin/python /path/to/overleaf-mcp/server.py
Claude Desktop — add to claude_desktop_config.json:
{
"mcpServers": {
"overleaf": {
"command": "/path/to/overleaf-mcp/.venv/bin/python",
"args": ["/path/to/overleaf-mcp/server.py"]
}
}
}
(Any other MCP client works the same way — run server.py with the venv's Python over stdio.)
5. Try it
Ask your assistant something like:
"List my Overleaf projects" "Fix the compile errors in my thesis project" "Update the abstract in main.tex and recompile"
How it works
- Authenticates with the
overleaf_session2cookie (same session your browser uses) - Reads/writes files through Overleaf's project API — edits show up live in the web editor and in project history
- Compiles run on Overleaf's servers, so you get the exact same output as clicking "Recompile"
Security notes
- Your session cookie grants full access to your Overleaf account — keep
.envprivate (it's gitignored) - The server runs locally over stdio; nothing is exposed to the network
License
MIT
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。