Configuring and Using Singdata Lakehouse MCP Server

Advanced Configuration

1. Enabling MCP Tool Integration

If your environment supports MCP (Model Context Protocol) tools, you can add an MCP server via the command line:

Add an HTTP-type MCP server:

Datus> .mcp add --transport http clickzetta_mcp_http http://localhost:8002/mcp

Or add an SSE-type MCP server:

Datus> .mcp add --transport sse clickzetta_mcp_sse http://localhost:8003/sse

View the list of added MCP servers:

Datus> .mcp list

Check MCP server connection status:

Datus> .mcp check clickzetta_mcp_http

Call an MCP tool (format: server_name.tool_name; check specific tool names via .mcp check):

Datus> .mcp call clickzetta_mcp_http.<tool_name>

Configure tool filters (optional):

Datus> .mcp filter set clickzetta_mcp_http --allowed tool1,tool2 --enabled true

View tool filter configuration:

Datus> .mcp filter get clickzetta_mcp_http

Remove an MCP server (if needed):

Datus> .mcp remove clickzetta_mcp_http

MCP Command Reference:

.mcp list - View all MCP servers
.mcp check <name> - Check server connection status and display available tools
.mcp call <server.tool> - Call a specific tool, format: "server_name.tool_name"
.mcp filter set <server> --allowed tool1,tool2 - Set the allowed tools list
.mcp filter get <server> - View current filter configuration
.mcp remove <name> - Remove an MCP server

Important Note: To find specific available tool names, first add the MCP server, then view them using the .mcp check <server_name> command.

Using MCP Tools via Subagent (Recommended)

After adding an MCP server, it is recommended to use MCP tools through a subagent, which allows you to select the desired tools during the addition process:

Add a subagent (launches interactive wizard):

Datus> .subagent add

View all configured subagents:

Datus> .subagent list

Build a knowledge base for a subagent (optional):

Datus> .subagent bootstrap agent_name

Remove a subagent (if needed):

Datus> .subagent remove agent_name

Subagent Command Reference:

.subagent add - Launch the interactive wizard to add a new intelligent agent; MCP tools can be selected during this process
.subagent list - List all configured subagents
.subagent bootstrap <agent_name> - Build a specialized knowledge base for the specified agent
.subagent remove <agent_name> - Remove the specified subagent

When adding a subagent, set the following:

agent_description: ClickZetta Lakehouse assistant with MCP tool integration

rules: - Detect user intent: distinguish between analysis requests and SQL generation requests - For analysis/explanation requests: focus on interpreting results, use markdown format - For SQL generation requests: provide executable statements with explanations in JSON format - Use MCP tools for instance switching, job history, system operations,etc - When users mention "instance switching", "job history", or similar operations - call MCP tools - Prefer specialized analysis tools over basic data collection tools when available - Use single comprehensive tools over multiple basic tool combinations when possible - Choose tools designed for the specific analysis type rather than generic tools - When using MCP analysis tools: focus on explaining the tool results rather than auto-generating SQL

Benefits of using MCP tools through the subagent approach:

Interactively select desired tools during the addition process
Create specialized intelligent agents for each scenario
Support building specialized knowledge bases to enhance analysis capabilities

Features After Enabling MCP Tools:

Instance management and switching
Job history queries
Advanced data analysis tools
System monitoring functions

Last updated: November 2025