How MCP Connects AI Tools

MCP (Model Context Protocol) is an open standard created by Anthropic that lets AI assistants connect to external tools and data sources. It is the protocol that makes cross-platform AI memory possible.

What MCP does

Without MCP, each AI assistant is an isolated island. Claude cannot access your database. ChatGPT cannot read your files. Cursor cannot call your APIs. Each tool has its own proprietary plugin system (or none at all).

MCP standardizes this. An MCP server exposes tools (functions the AI can call) and resources (data the AI can read). Any MCP client — Claude Code, Claude Desktop, Cursor, Codex CLI, Gemini CLI — can connect to any MCP server.

How HippoDid uses MCP

HippoDid’s MCP server exposes memory operations as tools:

Tool	What it does
add_memory	Store a fact in cloud memory
search_memories	Retrieve relevant facts by semantic search
sync_file	Upload a file to cloud storage
list_characters	List all memory namespaces
export_character	Export all memories as Markdown

When you connect HippoDid to Claude Code, Claude can call add_memory to store something and search_memories to recall it. The same server connected to Cursor gives Cursor the same capabilities — and the same memories.

Two ways to connect

Hosted (Streamable HTTP)

The simplest setup. Point your MCP client at https://api.hippodid.com/mcp with your API key. No local software needed.

{
  "hippodid": {
    "type": "http",
    "url": "https://api.hippodid.com/mcp",
    "headers": {
      "Authorization": "Bearer hd_sk_YOUR_KEY"
    }
  }
}

Local JAR (stdio)

For advanced features like background file watching. Download the MCP server JAR and run it locally.

{
  "hippodid": {
    "command": "java",
    "args": ["-jar", "/path/to/hippodid-mcp-server.jar"],
    "env": { "HIPPODID_API_KEY": "hd_sk_YOUR_KEY" }
  }
}

See MCP Setup for per-client configuration guides for Claude Code, Claude Desktop, Cursor, Gemini CLI, Codex CLI, ChatGPT, and OpenClaw.