For a hobby project I have created a Model Context Protocol (MCP) server that can be used for searching the content of my blog posts from an AI application.
The initial implementation is just a stdio MCP Java app that obtains the blog posts’ contents and holds it in a local index with embeddings to support semantic search.
The two tools exposed are search_posts and get_post.
I used Claude code to develop the MCP server, then tried out setting up some configuration for Claude CLI and a model in LM Studio.
The implementation is fine for an isolated small experiment, but wouldn’t be my choice of approach for a scalable production system.
Later on I may try out setting up an HTTP MCP service that would be closer to a useful scalable production system for a knowledgebase.
Screenshots of Claude Searching via MCP
The code
See the following GItHub repository:
Leave a Reply