@0latency/mcp-server

@0latency/mcp-server

Persistent memory for AI agents. Give Claude, Cursor, Windsurf, or any MCP-compatible tool long-term memory in one config change.

0Latency automatically extracts, stores, and recalls memories across conversations — so your AI assistant actually remembers you.

Features

• Automatic memory extraction — Detects facts, preferences, and relationships from conversations
• Semantic recall — Retrieves the right memories for the current context
• Knowledge graph — Tracks entities and relationships across all conversations
• Document & conversation import — Bulk-load context from existing docs or chat exports
• Token-budgeted recall — Control exactly how much context gets injected

Quick Start

1. Get an API Key

Sign up at 0latency.ai and grab your API key from the dashboard.

2. Configure Your AI Tool

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "0latency": {
      "command": "npx",
      "args": ["-y", "@0latency/mcp-server"],
      "env": {
        "ZERO_LATENCY_API_KEY": "your-api-key-here"
      }
    }
  }
}

Claude Code

Add to your project .mcp.json:

{
  "mcpServers": {
    "0latency": {
      "command": "npx",
      "args": ["-y", "@0latency/mcp-server"],
      "env": {
        "ZERO_LATENCY_API_KEY": "your-api-key-here"
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "0latency": {
      "command": "npx",
      "args": ["-y", "@0latency/mcp-server"],
      "env": {
        "ZERO_LATENCY_API_KEY": "your-api-key-here"
      }
    }
  }
}

Windsurf / Other MCP Clients

Same pattern — just point your client's MCP config at:

command: npx -y @0latency/mcp-server
env: ZERO_LATENCY_API_KEY=your-api-key-here

3. Alternative: Install Globally

npm install -g @0latency/mcp-server

Then use 0latency-mcp as the command instead of npx:

{
  "command": "0latency-mcp",
  "env": {
    "ZERO_LATENCY_API_KEY": "your-api-key-here"
  }
}

Available Tools

Environment Variables

How It Works

1. Your AI tool connects to the 0Latency MCP server via stdio
2. During conversations, the AI calls memory_add or remember to store important facts
3. In future conversations, the AI calls memory_recall to retrieve relevant context
4. Memories compound — the more you use it, the better your AI knows you

No prompt engineering required. The AI tools are designed to be self-describing — your AI assistant will naturally use them when appropriate.

Usage Examples

Store a memory:

"Remember that I prefer TypeScript over JavaScript and use Vim keybindings."

Recall memories:

"What do you know about my coding setup?"

Import existing context:

"Import this project README as background knowledge."

Explore the knowledge graph:

"What entities are connected to the auth-service project?"

Development

git clone https://github.com/0latency-ai/mcp-server.git
cd mcp-server
npm install
npm run build
npm start

Links

• 🌐 0latency.ai — Product & dashboard
• 📖 API Documentation — Full API reference
• 🐛 Issues — Bug reports & feature requests

License

MIT © Justin Ghiglia