Desktop Install

Run Crow locally on your personal computer. No cloud, no server — everything runs on your machine and connects directly to your AI tools via stdio transport.

What This Gives You

• All core Crow features: memory, projects, sharing, blog
• Direct connection to Claude Desktop, Claude Code, Cursor, Windsurf, Cline, and more
• Local SQLite database — your data stays on your machine
• No accounts or API keys needed for core features

Prerequisites

macOS

# Install Homebrew if you don't have it
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# Install Node.js and Git
brew install node git

Linux (Ubuntu/Debian)

# Install Node.js 20
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt install -y nodejs git

Windows

# Download and install Node.js from https://nodejs.org (LTS version)
# Git: download from https://git-scm.com/download/win
# Or use winget:
winget install OpenJS.NodeJS.LTS
winget install Git.Git

Verify your installation:

node --version   # Should be 18.x or later
npm --version
git --version

Install Crow

git clone https://github.com/kh0pper/crow.git
cd crow
npm run setup

This installs dependencies and creates a local SQLite database. No API keys or external services needed.

Generate MCP Configuration

npm run mcp-config

This creates .mcp.json with your local server paths. Only servers with the required environment variables are included — core servers (memory, projects, sharing, blog) are always included since they need no API keys.

Connect to Your AI Platform

Claude Desktop

npm run desktop-config

Copy the output JSON into your Claude Desktop config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

Restart Claude Desktop. Look for the MCP server icons in the input area.

Claude Code

Claude Code automatically reads .mcp.json from your project:

cd crow
claude

Cursor / Windsurf / Cline

These editors read .mcp.json from your project directory. Open the crow folder in your editor and the MCP servers will be available.

See the Platforms guide for detailed setup instructions for each AI tool.

Optional: Add External Integrations

Edit .env to add API keys for external services:

cp .env.example .env
# Edit .env with your preferred editor

After editing, regenerate the MCP config:

npm run mcp-config

See the Integrations page for all available services and where to get API keys.

Optional: Multi-Device Access with Tailscale

By default, your desktop install only works on the machine running the servers. To access Crow from your phone or other devices:

1. Start the gateway: node servers/gateway/index.js
2. Install Tailscale on this machine and your other devices
3. Access your Crow at http://<tailscale-hostname>:3001 from any device

See the Tailscale Setup Guide for details.

Reducing Context Usage

By default, each Crow server is a separate entry in your AI tool's config. For fewer context tokens, use the combined server:

npm run mcp-config -- --combined

This generates a single crow-core entry that activates servers on demand. See the Context & Performance guide for details.

Limitations

• Only accessible from this machine (unless you add Tailscale or the gateway)
• No web-based Crow's Nest (requires the gateway)
• No public blog (requires the gateway + HTTPS)
• For remote access from any device, see Oracle Cloud or Home Server