​
Overview

The Memory Protocol MCP Server implements the Model Context Protocol (MCP), allowing you to embed Memory Protocol tools and APIs into your own AI assistants like Claude Desktop, Claude Integrations and Cursor.

​
What is MCP?

MCP (Model Context Protocol) provides a standardized way for AI assistants to access external tools and data sources. Our implementation focuses on enabling AI assistants to query the Memory Identity API and Social Graph using any MCP enabled AI client.

Learn more at modelcontextprotocol.io

MCP Tools Query Examples

See how you can use Memory Protocol MCP tools to query data

​
Setup

The server is running remotely. You’ll use it by configuring your AI software of choice to connect to https://mcp.memoryproto.co.

​
Prerequisites

Either:

  1. Claude Integrations access

Or:

  1. An MCP Client Application (like Claude Desktop, Cursor, etc..)
  2. Node.js (>= v18.0.0) and npm: Download from nodejs.org. (because we’ll use mpc-remote via npx to convert the remote MCP SSE transport responses to STDIO for the local clients)

​
Client configuration

​
Claude Integrations

Claude Integrations are configured through the web dashboard (not in Claude Desktop). New tools can only be added by Team Owners on Team Accounts and Pro accounts. Once configured in the web dashboard by an account or team owner tools will be available on the web and in the Claude Desktop app.
  1. Visit https://claude.ai on the web
  2. Go to Settings, by clicking your username in the bottom left.
  3. Scroll down to Integrations and choose Add more
  4. Enter a name for the integration: Memory Protocol
  5. Enter the URL: https://mcp.memoryproto.co/sse
  6. Select Add.
  7. Try asking Claude about an Ethereum address or ENS name.

​
Claude Desktop

  1. Open Claude Desktop
  2. Go to Settings and Developer
  3. Choose Edit Config
  4. Open the claude_desktop_config.json file
  5. Add an entry in the object for Memory:
{
  "mcpServers": {
    "memory-protocol": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.memoryproto.co/mcp"
      ]
    }
  }
}
  1. Save the configuration and restart or refresh your MCP client application.
  2. Try asking Claude about an Ethereum address or ENS name.

​
Cursor

  1. Open Cursor
  2. Open Cursor settings CMD+Shift+P and type Cursor Settings
  3. Choose MCP from the left side menu
  4. Choose Add new global MCP Server in the top right of the window
  5. Add an entry in the config object for Memory:
{
  "mcpServers": {
    "memory-protocol": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.memoryproto.co/mcp"
      ]
    },
  }
}
  1. Save the configuration and restart or refresh your MCP client application.
  2. Try asking your model about an Ethereum address or ENS name.

​
MCP Inspector

It can be useful to isolate MCP tools for testing purposes outside of an MCP client, where you can test the tools as a human with chosen arguments instead of relying on an LLM.

You can run the inspector directly from NPM using NPX:

  1. npx @modelcontextprotocol/inspector

This will boot up the Inspector server, which you can visit in your browser.

  1. Configure the connection on the left:
  • Transport Type: SSE
  • URL: https://mcp.memoryproto.co/sse
  1. Click Connect
  2. Once connected, click List Tools in the middle pane.
  3. Select a tool, enter an appropriate argument eg. lookup_wallet and enter vitalik.eth.
  4. Select Run tool.

You can call all the endpoints here without needing any LLM in the loop.