Content
# Tool List
> Make your study easier — Superstar MCP Automation Server
[](https://www.npmjs.com/package/easystudy-mcp)
[](LICENSE)
**easystudy-mcp** is a MCP (Model Context Protocol) server that connects to the **Superstar** platform, allowing you to complete course learning tasks automatically with an AI assistant — view courses, submit assignments, auto-complete courses, download materials, all with natural language interaction.
8 tools cover daily operations on the Superstar platform, supporting Claude Code, Cowork, and any MCP-compatible client.
## Tool List
| Tool | Function | Description |
|------|------|------|
| `login` | Login to Superstar | Phone number + password (DES encryption) |
| `get_courses` | Get course list | Return browser operation steps + JS extraction script |
| `get_homeworks` | Get assignment list | Navigate → click assignment label → view assignment status |
| `get_homework` | View assignment content | Read questions, options, question type |
| `submit_homework` | Submit assignment answers | Auto-fill MCQs / short answer questions, one-click submit |
| `get_homework_score` | View grading results | Display score, correct answer comparison |
| `auto_study` | Auto-complete task points | Video/audio/quiz/PDF fully automated processing |
| `download_materials` | Download course materials | PPT/PDF/Word/Excel/video |
## Installation
### Method 1: npx (Recommended, Zero Installation)
Configure directly in Claude Code, Cowork, or other MCP clients:
```json
{
"mcpServers": {
"easystudy": {
"command": "npx",
"args": ["easystudy-mcp"]
}
}
}
```
npx will automatically download and cache on the first run, and then start directly from the cache.
### Method 2: Global Installation
```bash
npm install -g easystudy-mcp
```
Configure in MCP client after installation:
```json
{
"mcpServers": {
"easystudy": {
"command": "easystudy-mcp"
}
}
}
```
### Method 3: Local Installation
```bash
git clone https://github.com/Cindorolla95/easystudy-mcp.git
cd easystudy-mcp
npm install
```
```json
{
"mcpServers": {
"easystudy": {
"command": "node",
"args": ["/path/to/easystudy-mcp/src/index.js"]
}
}
}
```
## Claude Code Configuration
Edit `.claude.json` or `CLAUDE.md` in the project root directory:
```json
{
"mcpServers": {
"easystudy": {
"command": "npx",
"args": ["easystudy-mcp"]
}
}
}
```
After starting Claude Code, MCP tools are automatically loaded, and you can operate Superstar with natural language:
> "Login to Superstar account" → "Show me my courses" → "What assignments are there in Statistics?" → "Help me do the first assignment"
## Cowork Configuration
Add in Cowork's MCP settings:
```
npx easystudy-mcp
```
## Usage Process
### Typical Assignment Process
```
1. login(phone, password) → Login to Superstar
2. get_courses() → Get course list, get courseId and classId
3. get_homeworks(courseId, classId) → Get assignment list, get workId
4. get_homework(courseId, classId, workId) → View questions (question type, options)
5. submit_homework(courseId, classId, workId, answers) → Submit answers
6. get_homework_score(courseId, classId, workId) → View grading results
```
### Auto-Complete Course Process
```
1. login(phone, password) → Login
2. auto_study(courseId, classId, {
autoVideo: true, // Auto-play video (mute + accelerate)
autoQuiz: true, // Auto-complete chapter quiz
playbackRate: 2, // Video speed (up to 4x)
})
```
### Answer Format
```json
// Single/Multiple Choice — Option letters
{"0": "A", "1": "C", "2": "D"}
// True/False — true/false
{"0": "true", "1": "false"}
// Short Answer/Calculation — HTML text
{"3": "<p>Answer: Hypothesis testing steps are as follows...</p>"}
// Mixed use
{
"0": "A", // Single choice question 1, choose A
"1": "false", // True/False question 2, choose false
"2": "<p>Solution: μ=100, σ=15</p>" // Calculation question
}
```
## Working Principle
**easystudy-mcp** adopts a **browser operation plan** architecture:
- `login` tool uses HTTP + DES encryption to complete login directly
- Other 7 tools return **structured operation plans** (URL, steps, JS code fragments), which are executed by AI agents through Chrome DevTools MCP in the browser
This design is because Superstar pages are dynamically rendered by JS, and pure HTTP requests cannot obtain course lists, assignment content, and other data. The "operation plan" approach allows AI to understand what to do in each step and execute flexibly in the browser.
## Security
- Passwords are transmitted with DES encryption (consistent with Superstar Android App encryption method)
- Session cookies are stored only in MCP process memory
- No data is uploaded to third-party servers
- Auto-complete courses have muted + accelerated video playback to avoid interfering with normal use
## Running from Source
```bash
git clone https://github.com/Cindorolla95/easystudy-mcp.git
cd easystudy-mcp
npm install
node src/index.js
```
MCP protocol communicates through stdio, and the service waits for MCP client connection after startup.
## Python Version
The project also provides a Python version of the Superstar MCP server (`python/` directory). Suitable for users who prefer a Python environment.
```bash
cd python/cxmooc_mcp
pip install -e . --break-system-packages
cxmooc-mcp # Start MCP service
```
Python version MCP client configuration:
```json
{
"mcpServers": {
"easystudy": {
"command": "python",
"args": ["-m", "cxmooc"]
}
}
}
```
## License
Apache-2.0
<p align="center">
<sub>Let AI help you with Superstar ✨</sub>
</p>
