内容
# dianshu-mcp
[English](./README.md) | [简体中文](./README_zh.md)
> - GitHub: https://github.com/YeeZTech/dianshu-mcp
> - Gitee: https://gitee.com/YeeZTech/dianshu-mcp
MCP server for the [Dianshu Data Platform](https://dianshudata.com) — giving AI Agents full access to dataset trading, including login, order management, encrypted downloads, API calls, and marketplace search.
## Deployment
### Option A (Recommended): Install from Releases
1. Download the latest Release archive (choose your system):
- macOS Apple Silicon (M1/M2/M3): `macos-arm64.zip`
- macOS Intel: `macos-amd64.zip`
- Linux x86_64: `linux-amd64.zip`
- Linux arm64: `linux-arm64.zip`
- Windows x86_64: `windows-amd64.zip`
- Windows arm64: `windows-arm64.zip`
2. After extraction you'll get:
- Binary: `dianshu-mcp` (Windows: `dianshu-mcp.exe`)
- Skills directory: `skills/`
### Option B: Build from Source
Prerequisites: **Go 1.22+** (all platforms), **Git**
```bash
# Clone from GitHub
git clone https://github.com/YeeZTech/dianshu-mcp.git
# Or Clone from Gitee
git clone https://gitee.com/YeeZTech/dianshu-mcp.git
cd dianshu-mcp
go build -o dianshu-mcp .
```
### Launch
```bash
./dianshu-mcp -headless=true
```
| Flag | Default | Description |
|------|---------|-------------|
| `-headless` | `false` | `true` = background mode (no browser popup); `false` = foreground mode |
| `-port` | `18061` | HTTP listen port |
| `-output-dir` | `~/Downloads/dianshu-mcp/` | Override output directory |
Output paths:
- Downloaded files → `~/Downloads/dianshu-mcp/downloads/`
- API results → `~/Downloads/dianshu-mcp/api-data/`
### Import Skills & Configure MCP
Copy `.skill/dianshu/` to your agent's skills directory, then configure the MCP connection. Both steps are required.
#### Hermes
**Import Skills:** `cp -r .skill/dianshu ~/.hermes/skills/`
**Configure MCP (`~/.hermes/config.yaml`):**
```yaml
mcp_servers:
dianshu-mcp:
transport: streamable-http
url: http://localhost:18061/mcp
```
Or via CLI: `hermes config set mcp_servers.dianshu-mcp.transport streamable-http` and `hermes config set mcp_servers.dianshu-mcp.url http://localhost:18061/mcp`
#### Claude Code
**Import Skills:** `cp -r .skill/dianshu .claude/skills/`
**Configure MCP (`.claude/settings.json`):**
```json
{
"mcpServers": {
"dianshu-mcp": {
"type": "streamable-http",
"url": "http://localhost:18061/mcp"
}
}
}
```
> ⚠️ Claude Code uses `"type"` instead of `"transport"`.
#### Cursor
**Import Skills:** `cp -r .skill/dianshu .cursor/skills/`
**Configure MCP (`.cursor/mcp.json`):**
```json
{
"mcpServers": {
"dianshu-mcp": {
"transport": "streamable-http",
"url": "http://localhost:18061/mcp"
}
}
}
```
#### Trae / Trae Solo
**Import Skills:** `cp -r .skill/dianshu .trae/skills/`
**Configure MCP:** Open Settings → MCP → Manual Add, or edit `.trae/mcp.json`:
```json
{
"mcpServers": {
"dianshu-mcp": {
"transport": "streamable-http",
"url": "http://localhost:18061/mcp"
}
}
}
```
#### WorkBuddy
**Import Skills:** `cp -r .skill/dianshu .workbuddy/skills/`
**Configure MCP:** Click CodeBuddy Settings in sidebar → MCP → Add MCP, add JSON, or edit `.workbuddy/mcp.json`:
```json
{
"mcpServers": {
"dianshu-mcp": {
"transport": "streamable-http",
"url": "http://localhost:18061/mcp"
}
}
}
```
#### Augment Code
**Import Skills:** `cp -r .skill/dianshu .augment/skills/`
**Configure MCP (`.augment/mcp.json`):**
```json
{
"mcpServers": {
"dianshu-mcp": {
"transport": "streamable-http",
"url": "http://localhost:18061/mcp"
}
}
}
```
#### Windsurf
**Import Skills:** `cp -r .skill/dianshu .windsurf/skills/`
**Configure MCP (`.windsurf/mcp.json`):**
```json
{
"mcpServers": {
"dianshu-mcp": {
"transport": "streamable-http",
"url": "http://localhost:18061/mcp"
}
}
}
```
#### Tongyi Lingma
**Import Skills:** `cp -r .skill/dianshu .tongyi/skills/`
**Configure MCP:** Avatar → Personal Settings → MCP → + → Manual Add (SSE type, enter name + `http://localhost:18061/mcp`), or edit `.tongyi/mcp.json`:
```json
{
"mcpServers": {
"dianshu-mcp": {
"transport": "streamable-http",
"url": "http://localhost:18061/mcp"
}
}
}
```
#### VS Code / Cline
**Import Skills:** `cp -r .skill/dianshu .vscode/skills/`
**Configure MCP (project root `mcp.json` or Cline extension settings):**
```json
{
"mcpServers": {
"dianshu-mcp": {
"transport": "streamable-http",
"url": "http://localhost:18061/mcp"
}
}
}
```
### Get Started
After connecting your agent, say "Log in to Dianshu" to scan the QR code. Then use natural language:
- "Show my purchased data" → list all orders
- "Download task XXX" → auto-download and decrypt
- "Search weather datasets" → search the marketplace
- "Call the Xiaohongshu API" → invoke a purchased data API
---
## Troubleshooting
### MCP tools not appearing after installation?
Skills and MCP are two independent configurations — missing either causes failure:
1. Skills copied to agent's skills directory?
2. MCP config has `http://localhost:18061/mcp`?
3. `dianshu-mcp` server running?
4. Restart agent
If still not working, call `check_login_status` manually to verify.
---
## MCP Tools (19)
### Account & Login
| Tool | Description |
|------|-------------|
| `check_login_status` | Check Dianshu login status |
| `get_login_qrcode` | Get WeChat QR login image (displays in chat) |
| `wait_login` | Wait for QR scan completion (use after get_login_qrcode) |
| `open_login_browser` | Open browser for login (QR + password) |
| `set_token` | Manually save login token (browser login fallback) |
| `delete_cookies` | Clear session, switch account |
### Orders & Downloads
| Tool | Description |
|------|-------------|
| `list_orders` | List orders, filter by type / code |
| `list_downloads` | List purchased downloadable data products |
| `download_order` | Download and decrypt with task code |
| `list_purchased_apis` | List purchased data APIs |
| `get_api_detail` | Get API parameter details |
| `call_api` | Call a purchased API (auto encrypt/decrypt) |
### Dataset Search
| Tool | Description |
|------|-------------|
| `search_datasets` | Search Dianshu marketplace by keyword |
| `dataset_detail` | Get dataset details |
| `homepage_recommend` | Get homepage recommendations (popular / high-rated) |
| `my_datasets` | Get my published datasets |
### Profile & Wallet
| Tool | Description |
|------|-------------|
| `get_my_profile` | Get account profile |
| `get_my_wallet` | Get wallet balance |
| `list_wallet_transactions` | View transaction history |
---
## Project Structure
```
dianshu-mcp/
├── main.go # Entry point
├── server.go # Application container
├── routes.go # HTTP routes
├── mcp.go # MCP tool registration (19 tools)
├── config/config.go # Unified configuration
├── logger/logger.go # Unified logging
├── handler/handler.go # MCP handler layer
├── service/service.go # Business layer
├── dianshu/ # Dianshu HTTP client
│ ├── api.go # API endpoints
│ ├── auth.go # WeChat QR login
│ ├── browser.go # go-rod browser automation
│ ├── cookies.go # Cookie persistence
│ ├── types.go # Data types
│ └── dataset_types.go # Dataset types
├── pkg/ # Standalone SDK modules
│ ├── chain/ # On-chain ops (chain.go + signer.go)
│ ├── crypto/ # Crypto (ECDH + AES-CMAC + AES-GCM)
│ ├── kms/ # KMS integration
│ ├── pipeline/ # Download pipeline
│ └── sdk/ # Data API SDK
├── .skill/dianshu/ # Agent skills (1 main + 4 sub)
├── go.mod / go.sum
└── README.md
```
## Tech Stack
- Language: Go 1.22+
- Web framework: Gin
- MCP SDK: `github.com/modelcontextprotocol/go-sdk`
- Browser automation: `github.com/go-rod/rod`
- Cryptography: `github.com/decred/dcrd/dcrec/secp256k1/v4`
- Ethereum: `github.com/ethereum/go-ethereum`
## Development
```bash
go build -o dianshu-mcp .
go fmt ./...
go test ./...
```
## License
MIT
连接信息
猜你喜欢
暂无推荐服务器