
Markdown
io.github.ofershapPublished by GitHub account ofershapv1.0.1Updated Oct 27, 2026
Markdown MCP — search, extract sections, list headings, find code blocks.
Overview
Local MCP server that lets an assistant search and navigate markdown files on disk by headings, sections, code blocks and frontmatter.
- What you can do
- List all .md files in a directory recursively
- Full-text search across markdown files, case-insensitive, up to 50 results
- Extract a section by heading and list headings as a table of contents
- Find fenced code blocks, optionally filtered by language
- Parse YAML frontmatter metadata at the start of a file
- When to use it
- Useful when your project wikis, ADRs, onboarding guides or internal docs live as markdown files in a repo or on disk and you want the assistant to search and pull out structured parts rather than raw text. It is a structured alternative to a plain filesystem server for documentation work.
- Requirements
- Runs locally over stdio, typically started with npx mcp-server-markdown, so Node.js and npm are needed. No API keys or authentication; the assistant passes directory and file paths to the tools. Desktop MCP clients such as Claude Desktop, Cursor or VS Code Copilot.
Installation
In SourceWeft
- Open Markdown in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
mcp-server-markdown
[npm version] [npm downloads] [CI] [TypeScript] [License: MIT] [Agent Plugins]
Search, navigate, and extract content from local markdown files. Full-text search, section extraction, heading navigation, code block discovery, and frontmatter parsing.
Works with Claude Desktop, Cursor, VS Code Copilot, and any MCP client. Reads local
.mdfiles, no auth needed.
[Install in Cursor] [Add to VS Code]
[MCP server for searching and navigating markdown documentation]
Demo built with remotion-readme-kit
Why
Tools like Context7 are great for looking up library docs from npm, but they don't help with your own documentation. Project wikis, internal knowledge bases, architecture decision records, onboarding guides: they all live as markdown files in your repo or on disk. The filesystem MCP server can read those files, but it treats them as raw text. It doesn't understand headings, sections, or code blocks. This server does. Point it at a directory and your assistant can search across all your docs, pull out a specific section by heading, list the table of contents, or find every TypeScript code example in your knowledge base.
Tools
Quick Start
Cursor
Add to .cursor/mcp.json:
Claude Desktop
Add to claude_desktop_config.json:
VS Code
Add to user settings or .vscode/mcp.json:
Examples
- "Search all docs in ./docs for mentions of 'authentication'"
- "Show me the 'API Reference' section from README.md"
- "List all headings in CONTRIBUTING.md"
- "Find all TypeScript code blocks in the docs"
- "What's the frontmatter metadata in this file?"
- "Give me the table of contents for our architecture docs"
Agent Plugins
This repo is an Agent Plugins 1.0.0 package: plugin.json, portable mcp.json, and skills/ ship together with the MCP server.
For Cursor, clone the repo and copy or symlink it to ~/.cursor/plugins/local/mcp-server-markdown, then reload the window. Skills and MCP show up under Customize > Plugins.
The Cursor and VS Code install buttons above still work: they add the same npx -y mcp-server-markdown stdio server as manual JSON.
FAQ
What is mcp-server-markdown?
An MCP server that searches and navigates local .md files: full-text search, sections by heading, TOC, code blocks, and YAML frontmatter.
How is this different from Context7?
Context7 pulls published library docs from npm. This server indexes markdown already on your disk (wikis, ADRs, internal guides).
How is this different from the filesystem MCP?
Filesystem tools read raw files. This server understands headings, sections, fenced code, and frontmatter.
Can I install it as an Agent Plugin in Cursor?
Yes. Use ~/.cursor/plugins/local/mcp-server-markdown so the bundled markdown-search skill loads with the MCP config.
Do I need API keys?
No. Paths are local; the agent passes directory and file paths to the tools.
Development
See also
More MCP servers and developer tools on my portfolio.
Author
README built with README Builder
License
MIT © 2026 Ofer Shapira
Source: README.md at commit f845545
Tools
0Version history
1- v1.0.1LatestSep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts
