omnivox-mcp
MCP server for the Omnivox platform used by Quebec CEGEPs. Enables AI agents to access course schedules, grades, MIO messages, evaluations, and documents from Omnivox-based institutions.
README
Omnivox MCP
An MCP server that exposes Quebec college student portals (Omnivox/Lea) through MCP tools and a REST tool gateway.
It runs a persistent Puppeteer browser logged into Omnivox, executing requests through the site's own JavaScript — so challenge-response auth, cookies, and encoding are all handled natively.
Features
- 30 MCP tools — courses, grades, schedule, calendar, messaging (MIO), documents, assignments, college news
- Two transport modes — HTTP (default, Express server with MCP-over-HTTP + REST tool gateway) or stdio (for MCP clients)
- REST tool gateway — all MCP tools exposed as plain HTTP endpoints for non-MCP agents
- File downloads — direct binary or temporary browser-friendly links (15 min TTL)
- Internal messaging — read, search, send, flag, move, and delete MIO messages
- Persistent session — Chrome profile survives restarts, no repeated logins
Quick Start
# 1. Clone and install
git clone https://github.com/Beat-YT/omnivox-mcp.git
cd omnivox-mcp
npm install
# 2. Authenticate (first time only)
cd omnivox-connection
npm install && npm start
# Log in through the Electron window, then close it
# 3. Start the server
cd ..
npm start # HTTP mode (default) — Express server on port 3000
npm run start:stdio # stdio mode — for MCP clients
HTTP mode (default): Starts an Express server with MCP-over-HTTP at /mcp?key=... and a REST tool gateway. Access key is auto-generated at ~/.omnivox/accessKey.txt.
stdio mode: The MCP client launches the server as a subprocess and communicates over stdin/stdout. No Express server, no access key needed.
Each instance serves one Omnivox account. The server maintains a single browser session tied to the account you logged in with.
Re-authentication
If your session expires or you need to log into a different account, reset the session data and re-run the Electron app:
npm run reset
cd omnivox-connection && npm start
# Log in, wait for the success dialog, then close the window
cd .. && npm start
npm run reset deletes ~/.omnivox/browser/, cookies.json, and config.json so the server starts fresh on next launch.
Configuration
Set these as environment variables or in a .env file at the project root:
| Variable | Default | Description |
|---|---|---|
PORT |
3000 |
Server port |
OMNIVOX_DATA_DIR |
~/.omnivox |
Data directory for config, cookies, browser profile, and access key |
MCP_SERVER_URL |
(none) | Optional. Public base URL that enables download link generation (get-document-link / get-assignment-file-link). Set to your public domain (e.g. https://omnivox.example.com). |
Usage
MCP (for AI assistants)
stdio mode: Add the server to your MCP client config — it launches the server as a subprocess. Use npm run start:stdio. See AGENT_SETUP.md for connection config.
HTTP mode: Connect your MCP client to http://localhost:3000/mcp?key=YOUR_KEY via Streamable HTTP transport.
Either way, the AI gets access to all 30 tools — ask it about your grades, schedule, assignments, or messages in natural language.
REST Tool Gateway (HTTP mode)
For agents or apps that don't support MCP, all tools are exposed as plain HTTP endpoints:
# List all available tools
curl http://localhost:3000/tools -H "x-mcp-auth: YOUR_KEY"
# Get your courses
curl -X POST http://localhost:3000/tools/get-courses-summary \
-H "x-mcp-auth: YOUR_KEY" -d '{}'
# Check your grades
curl -X POST http://localhost:3000/tools/get-grades-summary \
-H "x-mcp-auth: YOUR_KEY" -d '{}'
# Read your inbox
curl -X POST http://localhost:3000/tools/get-mio-messages \
-H "x-mcp-auth: YOUR_KEY" -d '{}'
The tools endpoint accepts JSON bodies regardless of Content-Type header — no need to set it explicitly.
Build your own projects
The REST tool gateway isn't just for AI — it's a full Omnivox API you can use in any project:
- Grade dashboard — pull your grades into a custom web app or spreadsheet
- Schedule widget — display your weekly timetable on a home screen or desktop widget
- Assignment tracker — build notifications for new assignments or upcoming deadlines
- Discord/Slack bot — get pinged when new grades or messages come in
- Mobile app — build a custom Omnivox client with just the features you care about
- Data export — archive your grades, documents, and messages for your own records
Any language that can make HTTP requests works — Python, JavaScript, Swift, Kotlin, Go, whatever you prefer. Use GET /tools to discover all available tools and their input schemas.
Documentation
See the Wiki for full documentation:
- Setup — prerequisites, installation, environment variables
- Architecture — how the system is structured
- MCP Tools — all 30 tools with parameters
- REST Tool Gateway — using the tool gateway for non-MCP agents
- Authentication — access keys, web tokens, session management
Is this safe?
We never see your password or data. Everything runs locally on your machine (or your own server):
- You log in yourself through the Electron app — your Omnivox password is entered directly on the official Omnivox login page. This project never asks for, stores, or transmits your password.
- No external servers. The MCP server runs entirely on your machine. Your data never passes through any third-party service — all requests go directly between your server and Omnivox.
- Session stays local. The browser session and cookies are stored in
~/.omnivox/on your disk. Only you have access to them. - Open source. The entire codebase is here for you to inspect. There are no hidden network calls, analytics, or telemetry.
Security best practices
- Never expose the server without HTTPS. Use a reverse proxy (e.g. Nginx) with TLS. See the Setup wiki for a sample config.
- Keep
~/.omnivox/private. It contains your Omnivox session and access key. TreataccessKey.txtlike a password — do not share it with anyone. - See Security Precautions for more details.
Disclaimer
This project is unofficial and not affiliated with, endorsed by, or associated with Skytech Communications, Omnivox, or any Quebec college. It is provided as-is for personal and educational use only. Use it at your own risk — the authors are not responsible for any consequences resulting from its use, including but not limited to account restrictions, data loss, or violations of your institution's terms of service.
License
ISC
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。