Sage MCP
An academic research server that provides LLMs with full-text search and retrieval access to academic papers and periodicals. It enables research-grounded responses through features like topic filtering and programmatic database expansion.
README
Sage MCP - Academic Research Server
An MCP (Model Context Protocol) server that provides LLMs with access to academic papers and periodicals for research-grounded responses.
Features
- Full-text search across papers using SQLite FTS5
- Topic/publication filtering for targeted research
- Paper retrieval with full content for detailed analysis
- Extensible database - add your own papers programmatically
Tools Available
| Tool | Description |
|---|---|
search_papers |
Semantic/keyword search across all papers |
get_paper |
Retrieve full paper content by ID |
list_topics |
Browse available research topics |
list_publications |
Browse available journals/sources |
get_papers_by_topic |
Filter papers by topic |
get_papers_by_publication |
Filter papers by journal |
add_paper |
Add new papers to the database |
Quick Start (Local Setup)
Prerequisites
- Node.js 18+ (recommended: 20 LTS)
- npm 9+
1. Clone/Download and Install
cd C:\AIapps\SageMCP # or your chosen directory
npm install
npm run build
2. Seed the Database
# Option A: Sample papers (10 ML/AI papers)
npm run seed
# Option B: Import from XML (if you have papers_database.xml)
npm run import-xml
3. Verify Installation
npm start
# Should output: "Sage MCP Server running on stdio"
# Press Ctrl+C to stop
Connecting to MCP Clients
Claude Desktop (Windows)
-
Locate config file:
%APPDATA%\Claude\claude_desktop_config.jsonUsually:
C:\Users\<YourName>\AppData\Roaming\Claude\claude_desktop_config.json -
Edit the config (create if doesn't exist):
{ "mcpServers": { "sage": { "command": "node", "args": ["C:/AIapps/SageMCP/dist/index.js"] } } } -
Restart Claude Desktop completely (quit from system tray, reopen)
-
Verify connection: Look for the hammer 🔨 icon in Claude's chat input - click it to see available tools
Claude Desktop (macOS)
-
Config location:
~/Library/Application Support/Claude/claude_desktop_config.json -
Config content:
{ "mcpServers": { "sage": { "command": "node", "args": ["/Users/<you>/path/to/SageMCP/dist/index.js"] } } }
Windsurf / Cascade
Add to your MCP settings (Settings → MCP Servers):
{
"sage": {
"command": "node",
"args": ["C:/AIapps/SageMCP/dist/index.js"]
}
}
Cline (VS Code Extension)
- Open VS Code settings
- Search for "Cline MCP"
- Add server configuration:
{ "sage": { "command": "node", "args": ["C:/AIapps/SageMCP/dist/index.js"] } }
Custom MCP Client (Programmatic)
import { Client } from '@modelcontextprotocol/sdk/client/index.js';
import { StdioClientTransport } from '@modelcontextprotocol/sdk/client/stdio.js';
const transport = new StdioClientTransport({
command: 'node',
args: ['C:/AIapps/SageMCP/dist/index.js'],
});
const client = new Client({ name: 'my-client', version: '1.0.0' }, {});
await client.connect(transport);
// List available tools
const tools = await client.listTools();
console.log(tools);
// Call a tool
const result = await client.callTool({
name: 'search_papers',
arguments: { query: '神學', limit: 5 }
});
console.log(result);
Configuration
Environment Variables
| Variable | Description | Default |
|---|---|---|
SAGE_DB_PATH |
Custom database file path | ./data/sage.db |
Example with custom DB path:
{
"mcpServers": {
"sage": {
"command": "node",
"args": ["C:/AIapps/SageMCP/dist/index.js"],
"env": {
"SAGE_DB_PATH": "D:/Research/my_papers.db"
}
}
}
}
Adding Your Own Papers
Via the MCP Tool
LLMs can use the add_paper tool directly to add papers.
Via XML Import
Place your XML file in data/ and run:
npm run import-xml -- ./data/your_papers.xml
Programmatically
import { initDb, addPaper } from './dist/db.js';
await initDb();
addPaper({
title: 'Your Paper Title',
authors: 'Author One, Author Two',
abstract: 'Paper abstract...',
content: 'Full paper content...',
publication: 'Journal Name',
publication_date: '2024-01-15',
doi: '10.1234/example',
url: 'https://example.com/paper',
topics: 'Topic1, Topic2',
keywords: 'keyword1, keyword2, keyword3'
});
Database Schema
Papers are stored in SQLite with full-text search indexing:
| Field | Type | Description |
|---|---|---|
id |
INTEGER | Auto-incrementing primary key |
title |
TEXT | Paper title |
authors |
TEXT | Comma-separated author list |
abstract |
TEXT | Paper abstract |
content |
TEXT | Full paper text |
publication |
TEXT | Journal/conference name |
publication_date |
TEXT | Publication date (YYYY-MM-DD) |
doi |
TEXT | DOI identifier (optional) |
url |
TEXT | URL to paper (optional) |
topics |
TEXT | Comma-separated topics |
keywords |
TEXT | Comma-separated keywords |
VPS Deployment
Single command deploys everything (from your local machine):
# Windows PowerShell
.\deploy\deploy.ps1 -VpsHost "user@your-vps.com"
# Linux/macOS
./deploy/deploy.sh user@your-vps.com
This automatically:
- Installs Node.js and dependencies on VPS
- Creates
sageuser and directories - Copies source files and database
- Builds the project
- Sets up and starts systemd service
Uninstall:
# Windows
.\deploy\uninstall.ps1 -VpsHost "user@your-vps.com"
# Linux/macOS (or directly on VPS)
./deploy/uninstall.sh
See deploy/README.md for detailed options and troubleshooting.
Troubleshooting
"Cannot find module" errors
npm run build # Rebuild TypeScript
Claude Desktop doesn't show tools
- Check config JSON syntax (use a JSON validator)
- Ensure path uses forward slashes:
C:/path/to/filenotC:\path\to\file - Restart Claude Desktop completely (quit from system tray)
- Check Claude's logs:
%APPDATA%\Claude\logs\
Database errors
# Reset database
rm -rf data/sage.db
npm run seed # or npm run import-xml
Test MCP server manually
echo '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | node dist/index.js
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。