LightRAG MCP Server
MCP server for integrating LightRAG with AI tools. Provides a unified interface for interacting with LightRAG API through the MCP protocol.
README
LightRAG MCP Server
MCP server for integrating LightRAG with AI tools. Provides a unified interface for interacting with LightRAG API through the MCP protocol.
Description
LightRAG MCP Server is a bridge between LightRAG API and MCP-compatible clients. It allows using LightRAG (Retrieval-Augmented Generation) capabilities in various AI tools that support the MCP protocol.
Key Features
- Information Retrieval: Execute semantic and keyword queries to documents
- Document Management: Upload, index, and track document status
- Knowledge Graph Operations: Manage entities and relationships in the knowledge graph
- Monitoring: Check LightRAG API status and document processing
Installation
This server is designed to be used as an MCP server and should be installed in a virtual environment using uv, not as a system-wide package.
Development Installation
# Create a virtual environment
uv venv --python 3.11
# Install the package in development mode
uv pip install -e .
Requirements
- Python 3.11+
- Running LightRAG API server
Docker and Docker Compose
This repo ships a Dockerfile for the MCP server and a docker-compose.yml that runs both LightRAG and the MCP server together.
Quick start (Docker Compose)
docker compose up -d --build
docker compose ps
MCP endpoint: http://localhost:8000/
LightRAG API: http://localhost:9621/
Usage
LightRAG MCP server supports two MCP transport modes:
- stdio (default): run through an MCP client configuration file (
mcp-config.json) - streamable-http: run as a standalone HTTP server for remote MCP clients
Command Line Options
LightRAG API connection:
--host: LightRAG API host (default: localhost)--port: LightRAG API port (default: 9621)--api-key: LightRAG API key (optional)
MCP transport:
--mcp-transport: MCP transport (stdioorstreamable-http, default: stdio)--mcp-http-host: MCP HTTP host (default: 127.0.0.1)--mcp-http-port: MCP HTTP port (default: 8000)--mcp-http-path: MCP HTTP base/mount path (default: /)--mcp-http-stateless: Enable stateless HTTP mode (new session per request)--mcp-http-json-response: Return JSON responses instead of SSE for HTTP
Integration with LightRAG API
The MCP server requires a running LightRAG API server. Start it as follows:
# Create virtual environment
uv venv --python 3.11
# Install dependencies
uv pip install -r LightRAG/lightrag/api/requirements.txt
# Start LightRAG API
uv run LightRAG/lightrag/api/lightrag_server.py --host localhost --port 9621 --working-dir ./rag_storage --input-dir ./input --llm-binding openai --embedding-binding openai --log-level DEBUG
Setting up as MCP server (stdio)
To set up LightRAG MCP as an MCP server, add the following configuration to your MCP client configuration file (e.g., mcp-config.json):
Using uvenv (uvx):
{
"mcpServers": {
"lightrag-mcp": {
"command": "uvx",
"args": [
"lightrag_mcp",
"--host",
"localhost",
"--port",
"9621",
"--api-key",
"your_api_key"
]
}
}
}
Development
{
"mcpServers": {
"lightrag-mcp": {
"command": "uv",
"args": [
"--directory",
"/path/to/lightrag_mcp",
"run",
"src/lightrag_mcp/main.py",
"--host",
"localhost",
"--port",
"9621",
"--api-key",
"your_api_key"
]
}
}
}
Replace /path/to/lightrag_mcp with the actual path to your lightrag-mcp directory.
Running over Streamable HTTP (standalone server)
Use this when you need remote access or want to host the MCP server behind HTTP infrastructure:
uv run src/lightrag_mcp/main.py \
--mcp-transport streamable-http \
--mcp-http-host 0.0.0.0 \
--mcp-http-port 8000 \
--mcp-http-path /mcp \
--host localhost \
--port 9621 \
--api-key your_api_key
MCP clients should connect to: http://localhost:8000/mcp
Docker runtime configuration
Provide secrets like LIGHTRAG_API_KEY at runtime (not in the image):
docker build -t lightrag-mcp:local .
docker run --rm -it --env-file docker/mcp-example.env lightrag-mcp:local
Available MCP Tools
Document Queries
query_document: Execute a query to documents through LightRAG API
Document Management
insert_document: Add text directly to LightRAG storageupload_document: Upload document from file to the /input directoryinsert_file: Add document from file directly to storageinsert_batch: Add batch of documents from directoryscan_for_new_documents: Start scanning the /input directory for new documentsget_documents: Get list of all uploaded documentsget_pipeline_status: Get status of document processing in pipeline
Knowledge Graph Operations
get_graph_labels: Get labels (node and relationship types) from knowledge graphcreate_entities: Create multiple entities in knowledge graphedit_entities: Edit multiple existing entities in knowledge graphdelete_by_entities: Delete multiple entities from knowledge graph by namedelete_by_doc_ids: Delete all entities and relationships associated with multiple documentscreate_relations: Create multiple relationships between entities in knowledge graphedit_relations: Edit multiple relationships between entities in knowledge graphmerge_entities: Merge multiple entities into one with relationship migration
Monitoring
check_lightrag_health: Check LightRAG API status
Development
Installing development dependencies
uv pip install -e ".[dev]"
Running linters
ruff check src/
mypy src/
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。