mdplex.
MCP

Connect your agent to MdPlex

The MdPlex MCP server lets any MCP-compatible AI client read and write your documents directly. HTTP+SSE transport. No plugins. No wrappers.

Prerequisites

01 Generate your API token

  1. Go to Settings → API Tokens in your MdPlex account
  2. Click Create token
  3. Name it descriptively — e.g. Claude Code — main
  4. Set scope to read_write
  5. Copy the token immediately — it's only shown once

If your agent only needs to read documents (not write back), a read-scoped token is sufficient and limits exposure if the token is compromised.

02 Configure your MCP client

Claude Code

Add MdPlex to your project's .claude/settings.json or your global Claude Code settings:

{
  "mcpServers": {
    "mdplex": {
      "type": "http",
      "url": "https://app.mdplex.com/mcp/documents",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

Cursor

In Cursor's MCP settings panel, add a new server:

Other MCP clients

Any client supporting HTTP+SSE transport works with the same endpoint and auth header. Refer to your client's MCP documentation for the exact configuration format.

03 Verify the connection

Once configured, ask your agent to list your documents:

"List my MdPlex documents"

The agent will call list_documents and return your document list. If you see your documents, you're connected.

Available tools

Tool What it does Scope
list_documents List all your active documents read
get_document Fetch a document's full content by ID read
search_documents Search documents by keyword read
create_document Create a new document write
update_document Replace a document's content write
delete_document Delete a document permanently write

Agent write-back

The core capability of MdPlex is bidirectional — your agent can write documents back, not just read them. This turns MdPlex into a shared memory layer between you and your agent.

Basic workflow

  1. Create the document — Start with a context doc (e.g. project-context.md) via the web UI
  2. Agent reads it — At session start, the agent fetches it via get_document
  3. Agent updates it — At session end, the agent calls update_document to record decisions, open questions, current state
  4. You review — Document shows agent_updated status — skim before your next session

Instructing your agent

Add this to your CLAUDE.md for automatic context in every session:

## Context
- At session start: read MdPlex document ID 42 (project context)
- At session end: update that document with decisions made and questions resolved

Troubleshooting

Agent can't connect / no MdPlex tools available

Agent can read but not write

401 Unauthorized errors

Rate limit errors (429)

Get started free

Free tier includes MCP access. No credit card required.