使用说明
这是什么
这是测试 ClickHouse MCP服务器 和 MotherDuck DuckDB MCP服务器 两者皆有 硝酸银 和 谷歌ADK
示例提示
- 使用文件（“/tmp/mcp/data/*/*.log.gz”，“JSONEachRow”）作为数据源，您将找到从Linux系统捕获的多个事件的JSON压缩文件。
- 根据event_type查找并统计不同类型的事件。
- 然后根据auth-event_type找到SSH登录的次数。
- 最后，找到通过SSH登录的唯一用户的数量以及他们登录的源IP。
运行独立的DuckDB MCP服务器
mfranz@t490:~/tmp/duckdb-mcp$ uvx mcp-server-motherduck --db-path ./vector.db --transport stream --port 8000
[motherduck] INFO - 🦆 MotherDuck MCP Server v0.8.0
[motherduck] INFO - Ready to execute SQL queries via DuckDB/MotherDuck
[motherduck] INFO - Query result limits: 1024 rows, 50,000 characters
[motherduck] INFO - Query timeout: disabled
[motherduck] INFO - Starting MotherDuck MCP Server
[motherduck] INFO - Database client initialized in `duckdb` mode
[motherduck] INFO - 🔌 Connecting to duckdb database
[motherduck] INFO - ✅ Successfully connected to duckdb database
[motherduck] INFO - Registering handlers
[motherduck] INFO - MCP server initialized in http-streamable mode
[motherduck] INFO - 🦆 Connect to MotherDuck MCP Server at http://127.0.0.1:8000/mcp
[uvicorn] INFO - Started server process [759779]
[uvicorn] INFO - Waiting for application startup.
[motherduck] INFO - StreamableHTTP session manager started
[motherduck] INFO - MCP server started with StreamableHTTP session manager
[uvicorn] INFO - Application startup complete.
[uvicorn] INFO - Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
运行独立的ClickHouse MCP服务器
设置环境变量，这显然会监听所有接口
export CLICKHOUSE_ENABLED=false
export CHDB_DATA_PATH=data
export CHDB_ENABLED=true
export CLICKHOUSE_MCP_SERVER_TRANSPORT=http
export CLICKHOUSE_MCP_BIND_HOST=0.0.0.0
uv run --with mcp-clickhouse mcp-clickhouse
其他环境
LiteLLM使用 OLLAMA_API_BASE
export OLLAMA_API_BASE=$OLLAMA_HOST
参考文献
- https://github.com/ClickHouse/mcp-clickhouse
- https://clickhouse.com/blog/integrating-clickhouse-mcp
- https://www.youtube.com/watch?v=rvkOpKhPWvQ
