Skip to content

Getting Started

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

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

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

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

Then restart Claude Desktop.

Run this command in your terminal:

Terminal window
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:

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

Run this command in your terminal:

Terminal window
opencode mcp add covalence /Applications/Covalence.app/Contents/MacOS/cov-mcp

ChatGPT supports MCP via remote connectors. Covalence currently ships as a local macOS binary, so it isn’t yet reachable by ChatGPT’s connector model. Remote/network transport is on the Covalence roadmap — ChatGPT support will land with that.

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.

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.

Covalence can also auto-index files from watched folders on your Mac. File memories are searchable alongside your AI-stored memories but are read-only — to update them, edit the original file and Covalence re-indexes it automatically.

Configure watched folders in the Covalence app under Settings > Files.

  • 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
  • Under the Hood — How Covalence finds the right memory, from embeddings to hybrid search