Connect Claude, Cursor & VS Code to Markdown with MCP

md.rajkar.dev is an MCP server that gives your AI a memory. Connect Claude Code, Cursor, or VS Code and it can search, read, write, append to, version, and organize your Markdown docs — securely, scoped to your account.

The Model Context Protocol (MCP)is an open standard that lets AI tools call external services. md.rajkar.dev ships a hosted MCP server, so your AI can publish and manage Markdown pages for you — no copy-pasting, no manual steps. Ask Claude to "publish these release notes and give me the link" and it just happens.

What your AI can do

Once connected, the MCP server turns md.rajkar.dev into a memory layer for your AI. Every tool is scoped to your account, and your docs also show up as MCP resources you can @-mention and as ready-made prompts (like "start today's worklog").

search_docs — full-text search across everything you own, returning ranked snippets. The fastest way to recall a note.
list_docs / recent_docs / get_doc — browse (paginated, tag-filterable), see what changed this week, and read a doc
create_doc / upsert_doc — publish a page, or create-or-update by a deterministic slug (great for daily logs)
append_doc — atomically add to the end of a doc (one race-free call instead of read-modify-write) — ideal for worklogs
update_doc / delete_doc / move_docs — edit, rename, set tags or visibility, unpublish, or bulk-move into a project
list_versions / get_version / restore_version — full edit history and one-call undo
get_index — load your memory index (any doc you tag index); its contents are also injected automatically when your client connects
list_projects / create_project / list_project_docs / delete_project — organize docs into folders (deleting a folder keeps its docs)

Step 1 — Create an account and API key

Accounts are optional and need no email — just a username, a password, and a one-time recovery code. Create one, then open your account page and click New key. Give it a name (e.g. "Claude Code on my laptop") and copy the key — it starts with mdk_ and is shown only once.

Step 2 — Connect your client

The server speaks MCP over plain HTTP at https://md.rajkar.dev/api/mcp. Pick your client:

Claude Code

claude mcp add --transport http md https://md.rajkar.dev/api/mcp \
  --header "Authorization: Bearer mdk_your_key_here"

Cursor / Windsurf

Add this to your MCP config:

{
  "mcpServers": {
    "md": {
      "url": "https://md.rajkar.dev/api/mcp",
      "headers": { "Authorization": "Bearer mdk_your_key_here" }
    }
  }
}

VS Code

{
  "servers": {
    "md": {
      "type": "http",
      "url": "https://md.rajkar.dev/api/mcp",
      "headers": { "Authorization": "Bearer mdk_your_key_here" }
    }
  }
}

Any other MCP client works too — point it at the same URL with the same Authorization: Bearer header. The exact snippets, prefilled with your real key, are on the account page when you create a key.

Step 3 — Ask your AI

That's it. Try prompts like "list my docs", "create a doc titled Sprint Notes with this content…", or "move my changelog into the Releases project". Your AI calls the tools and hands you back live links.

Is it secure?

Yes. Each API key is tied to your account and only ever touches yourdocs — it can't see anyone else's. Keys are stored hashed, shown once, and you can revoke any key instantly from the account page. Your password is never shared with the AI. Anonymous publishing in the browser still works exactly as before — MCP is an optional add-on.

Ready? Create your API key and connect your first client.

Connect your AI to Markdown with MCP