Released: February 7, 2026
Version: desktop-v1.0.0
Platform: macOS (Apple Silicon)
Landing Page: next.henrybarefoot.com/engram
Download: GitHub Releases
Following the v1.1 release with deduplication, temporal queries, and the feedback loop, Engram now ships as a native macOS desktop application. No terminal required. No npm. Just download, drag to Applications, and your AI agents have persistent memory.
This release also marks the official npm publication as @hbarefoot/engram.
The npm package works great for developers comfortable with the terminal. But most people using Claude Desktop, Cursor, or ChatGPT aren't running npm install -g commands. They want something that just works.
The desktop app solves this:
.app.zip, unzip, drag to ApplicationsYou now have two ways to install Engram:
Engram_x.x.x_arm64.app.zip from GitHub ReleasesEngram.app to your Applications folderRequirements: macOS 12+ on Apple Silicon (M1/M2/M3/M4)
npm install -g @hbarefoot/engram
engram start
Requirements: Node.js 20+
Both options give you the same core functionality: MCP server, REST API, web dashboard, and all 6 memory tools.
The desktop app uses a modern hybrid architecture:
Tauri Shell (Rust)
├── React Frontend (Dashboard UI)
└── Node.js Sidecar (Engram Core)
├── MCP Server (port configurable)
├── REST API (localhost:3838)
└── SQLite Database (~/.engram/)
Tauri gives us a tiny native shell (~15MB vs Electron's ~150MB) with full macOS integration — menu bar, notifications, global shortcuts.
Node.js Sidecar lets us reuse 100% of the existing Engram codebase. No rewrite. The sidecar is the exact same code that runs via npm install -g @hbarefoot/engram, just bundled with its runtime.
The app includes everything needed to run offline:
| Component | Size | Purpose |
|---|---|---|
| Node.js binary | ~45MB | JavaScript runtime |
| Engram bundle | ~100KB | Core memory engine (esbuild CJS) |
| Native modules | ~15MB | better-sqlite3, onnxruntime-node |
| Embedding model | ~23MB | all-MiniLM-L6-v2 for semantic search |
Total app size: ~90MB — reasonable for a self-contained AI tool.
The desktop app can auto-configure these MCP-compatible agents:
| Agent | Config Method | Config Location |
|---|---|---|
| Claude Desktop | Auto | ~/Library/Application Support/Claude/claude_desktop_config.json |
| Claude Code | Auto | ~/.claude/mcp.json |
| Cursor | Auto | ~/.cursor/mcp.json |
| Windsurf | Auto | ~/.windsurf/mcp.json |
| ChatGPT | Manual | In-app Settings → Developer → MCP Servers |
ChatGPT now supports MCP servers, but configuration happens inside the app rather than a JSON file. The desktop app detects ChatGPT and provides setup instructions:
Engram lives in your menu bar, not the dock. Click the icon to:
The same React dashboard from the web version, but native:
On first launch, Engram scans for installed AI agents and offers to configure them. No JSON editing required. It:
The desktop app uses a custom build pipeline:
.node and .dylib files.app bundleThe Rust shell manages the Node.js sidecar:
All data lives in ~/.engram/:
~/.engram/
├── memory.db # SQLite database
├── config.json # User preferences
└── models/ # Cached embedding model
This is shared between the desktop app and npm package — you can switch between them without losing memories.
With the desktop app shipped, here's the updated roadmap:
| Version | Status | Features |
|---|---|---|
| v1.0 | ✅ Shipped | CLI, MCP server, REST API, web dashboard, agent auto-discovery |
| v1.1 | ✅ Shipped | Deduplication, temporal queries, feedback loop, context export |
| v1.0 Desktop | ✅ Shipped | Native macOS app, menu bar, bundled runtime |
| v1.2 | 🔜 Next | Windows/Linux desktop, system tray polish, PWA support |
| v2.0 | 📋 Planned | Graph relationships, team shared memory, Obsidian import |
Visit the Engram Landing Page for a visual overview and quick download links.
Desktop users:
npm users:
npm install -g @hbarefoot/engram
engram start