Quick Start

Get Cortex memory running with your AI agent in under a minute.

Automatic Setup

The fastest way to get started. This command will open the dashboard to create an API key, detect your installed editors (Claude, Kiro, Cursor), and configure them automatically.

npx -p cortex-memory-mcp cortex-memory-init

If you already have an API key, pass it directly:

npx -p cortex-memory-mcp cortex-memory-init ctx_your_key_here

OAuth Setup (No API Key)

Any MCP client that supports remote servers can connect via OAuth — no API key required. Just point it at the URL and sign in with Google or GitHub on first use.

{
  "mcpServers": {
    "cortex": {
      "url": "https://cortex-mem.com/api/mcp"
    }
  }
}

Tokens are cached for 30 days — no re-login between sessions.

Manual Setup

Prefer to configure things yourself? Follow these steps:

1. Create an API Key

Go to the Dashboard → API Keys page and generate a new key. Keys start with ctx_.

2. Add the MCP Server

Add the following to your AI client's MCP configuration:

{
  "mcpServers": {
    "cortex-memory": {
      "command": "npx",
      "args": ["cortex-memory-mcp"],
      "env": {
        "CORTEX_API_KEY": "ctx_your_key_here"
      }
    }
  }
}

3. Start Using Memory

Your agent now has three tools available: save_memory, search_memory, and delete_memory. It will automatically save important context and recall it when relevant.

What Gets Saved?

The agent decides what to remember based on the steering instructions bundled with the MCP server. Typical examples:

  • User preferences (editor, language, style)
  • Project architecture decisions
  • Debugging solutions and root causes
  • Deployment workflows and procedures
  • Environment details and configurations