sjtu

This MCP server provides a centralized gateway to SJTU campus services by exposing a set of tools that let the model query student information systems, email, venue reservations, and other campus portals through MCP-compatible interfaces. It integrates several native and Python plugins to enable actions such as fetching personal course schedules, viewing grades, reading or sending emails, and booking venues, all accessible via natural language commands. To use it, connect to the MCP server from your client (e.g., Cherry Studio or another MCP client) and enable the desired tool groups. You can steer the agent to perform a sequence of actions (for example, check your personal timetable, then fetch a specific exam date) using the exposed tool set and the MCP protocol.

The server supports two primary interaction modes: a standard HTTP/SSE-based stream and a stdio-based mode for local clients. When running in HTTP mode, you can connect with an endpoint such as http://localhost:5173/sse or http://localhost:5173/ for streamable HTTP, depending on configuration. Tools are grouped by domain (academic information, email, venue reservations, and Python plugins for additional campus services), and each tool exposes a defined function signature that the MCP model can call and interpret.

Prerequisites:\n- A machine with an OS compatible with the SJTU MCP server release (Windows is explicitly supported in the Quick Start).\n- Access to the internet to download the latest release from GitHub.\n- (Recommended) Python environment for optional plugin dependencies.\n\nInstallation steps:\n1) Download the latest release from the project GitHub Releases page.\n2) Unzip the package. You should find the executable (Windows) named SJTUGeek.MCP.Server.exe and related files.\n3) If you plan to use optional Python plugins, install Python 3.x and dependencies:\n\n pip install -r requirements.txt\n4) Start the MCP server according to your preferred mode:\n - For standard client connections (stdio/stdin-based): run the executable directly and pass any required flags (e.g., --cookie for authenticated tools) as described in the Quick Start.\n - For HTTP/SSE or Streamable HTTP: start the server with --use-http and provide cookie value if needed, then connect via the specified endpoint (e.g., http://localhost:5173/sse).\n5) In your MCP client (e.g., Cherry Studio), add a new server using the full path to the executable and configure arguments so that each line in the argument field represents a separate parameter (do not enclose in quotes). If you plan to enable Python plugins, include the appropriate DLL or identifier (e.g., python310.dll) as part of the arguments.\n6) If authentication is required for tool access, follow the Authorization steps in the README to obtain JAAuthCookie and pass it to the server using the --cookie option.

Tips and miscellaneous information:\n- For tools that require login-only access, you must pass JAAuthCookie via the --cookie parameter when starting the server. The Authorization section of the README explains how to obtain this cookie.\n- When using Python plugins, ensure Python is installed and accessible in your environment; some tools depend on Python execution for functionality.\n- If you plan to run the server as a service or in a container, you may adapt the command line and environment variables accordingly (e.g., mapping cookies and port exposure).\n- In the Cherry Studio client, add each tool as a separate entry under the server configuration, using the full executable path and one parameter per line, without quotes. For Python-based tools, ensure the Python runtime path or DLLs are correctly referenced as required by the plugin integration.\n- If you encounter connection issues, verify that the server is running with --use-http if you intend to use HTTP-based clients, and ensure the cookie value is correct and not expired.