kwdb

The KWDB MCP Server provides a protocol bridge for interacting with the KWDB database using the MCP (Model Context Protocol). It supports reading, writing, querying, and DDL operations, and automatically applies a LIMIT 20 to SELECT queries without a LIMIT clause to prevent large result sets. You can access the server via MCP tools either through standard IO (StdIO) or HTTP SSE depending on how the chosen runtime is exposed. The server exposes MCP Resources such as product_info, db_info, and table schemas, and offers two primary MCP Tools: read_query for read-only operations (SELECT, SHOW, EXPLAIN, etc.) and write_query for data modification and DDL operations (INSERT, UPDATE, DELETE, CREATE, ALTER, DROP). Prompts templates are embedded for guiding queries and workflows. When you run it in Docker, you can connect your MCP client to the container’s MCP endpoints and issue tools requests in the MCP JSON-RPC style, receiving standardized responses with content/results and explicit error signals.

Prerequisites:

• Docker installed and running on your host machine
• Basic familiarity with MCP clients and JSON-RPC style requests

Installation steps (Docker):

1. Pull the KWDB MCP Server Docker image (if not already available): docker pull kwdb/kwdb-mcp-server:latest
2. Run the MCP server in a container: docker run -i --rm kwdb/kwdb-mcp-server:latest
3. Connect an MCP client to the running server using the MCP protocol (StdIO or HTTP SSE, depending on your setup). You can also expose ports if the image is configured for HTTP endpoints (adjust docker run accordingly, e.g., -p 8080:8080).

Alternative (Building from source):

• Prerequisites: Go toolchain, Git
• Clone the repository and build the server binary: git clone https://github.com/kwdb/kwdb-mcp-server.git cd kwdb-mcp-server go build -o kwdb-mcp-server ./...
• Run the binary locally: ./kwdb-mcp-server

Notes:

• If you use Docker, ensure the image tag matches the version you intend to run.
• When exposing the server, configure appropriate network settings and security (firewalls, auth if provided by the server).

Tips and common considerations:

• The MCP Tools are divided into read-query (read-only) and write-query (DML/DDL). Use read_query for data retrieval to benefit from automatic LIMIT handling.
• Error handling follows standard MCP/JSON-RPC conventions with isError flags for tool results and jsonrpc errors for resource-related issues.
• If you enable HTTP SSE endpoints, you can stream query results and progress updates to MCP clients.
• Ensure the environment where the server runs has access to the KWDB database or service layer it targets; the server will validate queries and return appropriate RESOURCE_NOT_FOUND or INTERNAL_ERROR errors when resources or connections are unavailable.
• The server exposes MCP Resources in the kwdb:// namespace, such as kwdb://db_info/{database_name} and kwdb://table/{table_name}, which you can reference in prompts or client workflows.
• Review the embedded MCP Prompts (ReadExamples.md, WriteExamples.md) for ready-to-run example workflows you can adapt in your client integration.