Sparky runs on Windows, macOS, and Linux. Install it with a single command:

Or if you prefer a manual install:

• Download the latest release from GitHub
• Extract to your preferred location
• Add to your PATH environment variable

Requirements

• Windows 10/11, macOS 12+, or Ubuntu 20.04+
• Node.js 18+
• Git (for repository operations)

Once installed, open your terminal and run:

This launches the Sparky interface. From here you can:

• Ask questions about your codebase
• Request changes to files
• Run verification commands
• Spawn subagents for parallel work

First Task

Try asking Sparky to explain a file:

Sparky responds to natural language, but also supports these slash commands:

Session Control

• /new - Start a new session
• /compact - Compress conversation context
• /quit - Exit Sparky

Configuration

• /model - Change AI model
• /settings - Open settings
• /login - Sign in to your account
• /logout - Sign out

Tools

• /browser - Launch browser tools
• /memory - Search saved memories
• /subagents - View running subagents

For the best results with Sparky, follow this pattern:

1. Frame the Goal

Be specific about what you want:

• ❌ "Fix the login"
• ✅ "Add password validation to the login form that requires at least 8 characters"

2. Provide Context

Tell Sparky:

• Which files are relevant
• Any constraints or requirements
• What NOT to change

3. Review Changes

Sparky will propose edits. Review them carefully before accepting. You can:

• Ask for explanations
• Request modifications
• Reject changes you don't want

4. Verify

Always run verification commands:

Sparky can use these tools automatically when needed:

File Operations

• read - View file contents
• edit - Make precise file changes
• write - Create new files

Search

• grep - Search file contents
• bash - Run terminal commands
• web_search - Search the web

Browser

• browser_use - Read and interact with web pages

Spawn isolated workers for parallel tasks:

Each subagent runs independently with its own context. Results are merged back to the main thread.

Use Cases

• Code review (security, style, tests)
• Research multiple approaches
• Generate documentation
• Parallel file conversions

Your Code

• Source code is never used to train AI models
• Files are processed only to provide responses
• Data is encrypted in transit

Best Practices

• Never paste secrets or API keys in prompts
• Review all code changes before applying
• Use /logout when done on shared machines
• Keep Sparky updated for security patches

Sparky won't start

Check Node.js version:

Login issues

• Clear browser cookies for sparky.llc
• Run /logout then /login
• Check your internet connection

Context too long

Use /compact to summarize the conversation, or start a fresh session with /new.

Still stuck?

Contact support or check the changelog for known issues.