gkoos/confluence2md-mcp

This project is a small local server that lets AI coding assistants, such as VS Code with MCP support, Claude Code, or Codex CLI, search through a company's Confluence documentation. It uses a protocol called MCP, which is a standard way for AI tools to connect to external data sources and actions. This server does not talk to Confluence directly. Instead, it is the last step in a three part pipeline: one tool called confluence2md exports Confluence pages into markdown files, a second tool called confluence2md-indexer builds a searchable SQLite database from those files, and this project, confluence2md-mcp, serves that database to AI clients through MCP. To use it, you first need that SQLite index already built by the earlier tools in the pipeline, since this server cannot work with other data formats. Once you have the index, you download a prebuilt binary for your operating system and place it somewhere on your system path. The README gives configuration examples for setting it up in VS Code, Claude Code, and Codex CLI, each pointing the tool at the path of the index database through an environment variable. The server exposes one main tool, called confluence.search, which takes a search query along with several optional settings, such as choosing between hybrid, lexical, or vector search modes, filtering results by Confluence space or page, and setting date ranges. It returns ranked results along with score details for each matching chunk of content. The README also includes notes for developers who want to build the project from source using Go, along with a short troubleshooting section covering missing search results, Windows Subsystem for Linux path issues, and MCP clients that need a restart after the server is first registered.