ide integration

IDE Integration: Claude Desktop, Cursor, and Antigravity

Engram boots natively as a Model Context Protocol (MCP) background service. Instead of forcing you to memorize CLI commands or setup API endpoints, Engram exposes its memory tools natively directly to your LLM agent.

How it works

The MCP interface runs perfectly over standard operating system pipelines (Stdio). When you boot your IDE, it launches the engram binary silently in the background.

Whenever you type something like "Hey Claude, how does the user login script work?", the Agent autonomously fires the mcp_engram_recall tool inside its tool-chest, querying your .leg3 NVMe geometry map instantly.

Supported Tools

As of Phase 2, the Engram background daemon exposes the following explicitly Agent-executable tools:

• mcp_engram_remember: Embed and lock text into memory natively.
• mcp_engram_recall: Perform semantic extraction and K-NN bounding box sweeps.
• mcp_engram_watch_workspace: Commands the Daemon to attach an OS-level file-watch thread against a codebase.
• mcp_engram_pin: Prevent a critical context from being garbage collected by Autophagy.
• mcp_engram_remember_solution: Creates a perfectly crystallized Bug/Resolution tag linked mathematically inside the graph to stop future agents from failing the same compilation.
• mcp_engram_relate: Link nodes geometrically via OP_BIND circular convolutions.
• mcp_engram_context_for_file: Actively pre-load the highest correlation file contexts when an Agent opens a new .rs or .ts script.

Setup Instructions

1. Install Engram globally

First, confirm Engram is built and mapped to your system $PATH:

cargo install engram --git https://github.com/staticroostermedia-arch/engram

Note: Make sure ~/.cargo/bin is in your environment PATH.

2. Configure Claude Desktop

Add these blocks to your ~/.config/claude-desktop/claude_desktop_config.json:

{
  "mcpServers": {
    "engram": {
      "command": "engram",
      "args": ["mcp", "--store", "~/.engram/manifold"]
    }
  }
}

Restart Claude Desktop. The Engram tools will automatically appear via the hammer tooling icon.

3. Configure Google Antigravity

Inside the Google Deepmind Antigravity environment, Engram fits in natively via the mcpServers block located in your AppData directory (~/.gemini/antigravity/mcp_config.json):

{
  "mcpServers": {
    "engram": {
      "command": "engram",
      "args": ["mcp", "--store", "~/.engram/manifold"]
    }
  }
}

Reload the IDE window. The agent now possesses total persistence across sessions!