mcp-server-markdown is an MCP (Model Context Protocol) server specifically designed for searching, navigating, and extracting the contents of local Markdown files. Its core goal is to turn .md documents on disk into data that assistants can understand structurally, rather than just raw text.
[Purpose] Documents such as project wikis, internal knowledge bases, Architecture Decision Records (ADRs), and getting-started guides are often stored as Markdown files in repositories or on disk. Although a filesystem MCP can read these files, it treats them as plain text and does not understand headings, sections, or code blocks. mcp-server-markdown, by contrast, can recognize these structures, allowing assistants to search across all documents, extract specific sections by heading, list tables of contents, or find all TypeScript code examples in a knowledge base. The author also distinguishes it from Context7: Context7 is used to find published library documentation from npm, whereas this server indexes Markdown documents that already exist on disk.
[Main capabilities]
- Full-text search: perform full-text searches in Markdown files under a specified directory.
- Section extraction: extract specific sections by heading.
- Heading navigation: list all headings in a file and generate a table of contents (TOC).
- Code block discovery: find code blocks in documents, for example, find all TypeScript code blocks.
- frontmatter parsing: read the YAML frontmatter metadata of a file.
[Usage] This server can be connected as an MCP server to Claude Desktop, Cursor, VS Code Copilot, and any MCP client. Installation methods include adding configuration in .cursor/mcp.json, claude_desktop_config.json, or user settings and .vscode/mcp.json. It runs the npx -y mcp-server-markdown stdio server. The repository is also an Agent Plugins 1.0.0 package, containing plugin.json, a portable mcp.json, and skills/. For Cursor, you can clone the repository and copy or symlink it to ~/.cursor/plugins/local/mcp-server-markdown, then reload the window, and the skills and MCP will appear under Customize > Plugins. Usage examples include: searching for content about authentication in all documents under ./docs, displaying the API Reference section in README.md, listing all headings in CONTRIBUTING.md, finding all TypeScript code blocks in the documentation, viewing a file's frontmatter metadata, or getting the table of contents of an architecture document.
[Suitable scenarios] Suitable for users who need to let AI assistants understand and retrieve local Markdown document collections, such as developers consulting documentation in a repository, maintaining an internal knowledge base, or handling architecture decision records and getting-started guides.
[Limitations] No API key is required; paths are local paths, and the agent passes directories and file paths to the tools. Information about fees, pricing, platform support scope, and specific results is not provided in the material.
