Getting Started

import { Tabs, TabItem } from ‘@astrojs/starlight/components’;

Connect Covalence to your AI client by adding it as an MCP server. Once connected, your AI gains persistent memory that persists across sessions.

Prerequisites

• macOS 15 or later (Sonoma)
• Covalence installed in /Applications

Connect Your AI Client

Add the following to `~/Library/Application Support/Claude/claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "covalence": {
      "command": "/Applications/Covalence.app/Contents/MacOS/cov-mcp"
    }
  }
}
```

Then restart Claude Desktop.

Run this command in your terminal:

```bash
claude mcp add covalence /Applications/Covalence.app/Contents/MacOS/cov-mcp
```

Claude Code picks up the new server immediately — no restart needed.

Open **Cursor Settings > MCP Servers** and add the same JSON as Claude Desktop:

```json
{
  "mcpServers": {
    "covalence": {
      "command": "/Applications/Covalence.app/Contents/MacOS/cov-mcp"
    }
  }
}
```

Run this command in your terminal:

```bash
opencode mcp add covalence /Applications/Covalence.app/Contents/MacOS/cov-mcp
```

ChatGPT does not currently support the MCP protocol. Watch for updates from OpenAI.

Add the AI Instruction

Connecting the MCP server gives your AI access to the memory tools, but it won’t use them automatically without guidance. Add the AI Instruction to your AI client’s system prompt so it knows when to search and store.

This step is recommended for best results.

Verify It Works

Ask your AI: “What do you know about me?”

If Covalence is connected, it will use memory_search to look up relevant context and respond. On a fresh install the database is empty, so the AI will say it has no memories yet — that’s expected.

What’s Next

• Spaces — Isolate memories by project using the --space flag
• Core Memories — Pin important knowledge so your AI always checks it first
• AI Instruction — Full instruction text and where to add it per client