AI models are incredibly smart, but they suffer from one massive limitation: they don’t know who you are.

When you open a new session in Claude or ChatGPT, you start with a blank slate. The AI has no access to your project ideas, your meeting summaries, or the notes you took yesterday. You have to manually paste context, which is slow and repetitive.

With the release of the Model Context Protocol (MCP), this limitation is gone.

WorkingMemory speaks MCP natively. By setting up the WorkingMemory MCP server, you give Claude a secure, read-write bridge to your personal memory graph.

Here is how to set it up on your machine in under five minutes.

What is Model Context Protocol?

MCP is an open standard developed by Anthropic that allows local or remote client applications (like Claude Desktop) to securely expose data and tools to AI models.

Instead of building fragile, custom extensions for every app, developers build MCP servers. The AI client connects to these servers, discovers what tools are available, and runs them as needed.

Our server exposes two primary tools to your AI client:

  1. search_memory: Search your entire history of voice and text notes using natural language.
  2. save_thought: Quickly capture a new note or task directly from your chat session.

Setting Up the MCP Server

Step 1: Install Claude Desktop

If you haven’t already, download and install Claude Desktop (currently available for macOS and Windows).

Step 2: Grab Your API Token

  1. Open the WorkingMemory Web App.
  2. Go to Settings > Integrations > Developer Portal.
  3. Generate a new Personal Access Token and copy it to your clipboard.

Step 3: Configure Claude Desktop

Open your Claude Desktop configuration file. On macOS, this file is located at:

~/Library/Application Support/Claude/claude_desktop_config.json

On Windows, it is located at:

%APPDATA%\Claude\claude_desktop_config.json

Edit the file to include the working-memory server under the mcpServers block:

{
  "mcpServers": {
    "working-memory": {
      "command": "npx",
      "args": [
        "-y",
        "@workingmemory/mcp-server"
      ],
      "env": {
        "WORKING_MEMORY_TOKEN": "YOUR_PERSONAL_ACCESS_TOKEN"
      }
    }
  }
}

Make sure to replace "YOUR_PERSONAL_ACCESS_TOKEN" with the token you copied in Step 2.

Step 4: Restart Claude Desktop

Quit Claude Desktop entirely and launch it again.

You should see a new hammer icon in the bottom-right corner of the chat input box. Hovering over it will display the active tools provided by WorkingMemory: search_memory and save_thought.

Putting It to Work

Now, Claude has a direct line to your cobrain. Try asking questions that draw on your history:

“What did I mention wanting to talk to Colin about next time we grab dinner?”

Claude will automatically call search_memory, find the voice notes you recorded while walking the dog last Tuesday, and answer:

“You mentioned wanting to talk to Colin about Han Dynasty’s back room and his thoughts on Sichuan food after 9 PM.”

You can also save new details in real-time:

“Save a note that Colin’s ears are sensitive to loud spaces.”

Claude will call save_thought, and it will immediately sync to your phone’s WorkingMemory timeline.

By connecting your notes directly to your workspace, you stop wasting time copy-pasting your thoughts, and let your AI do the heavy lifting of remembering.