LocalCode

Turn natural language into CLI commands using Apple's on-device AI.

What is this?

LocalCode is a proof-of-concept that demonstrates building CLI tools with Apple's Foundation Models framework. Tell it what you want in plain English, it suggests the right command, you approve, and it runs.

All AI processing happens locally on your Mac. No cloud, no data leaving your machine.

Requirements

• Apple Silicon Mac (M1/M2/M3/M4)
• macOS 26+
• Xcode 26+ (for building the Swift helper)
• Go (for building the TUI)

Quick Start

git clone https://github.com/localcodeai/localcode.git
cd localcode/LocalCode

# Build the Swift AFM helper
cd Sources/afmhelper
swiftc -o afmhelper main.swift -framework FoundationModels -target arm64-apple-macosx26.0

# Build the Go TUI
cd ../..
go build -o localcode .

# Run
./localcode

How It Works

You: "list all python files"
LocalCode: find . -name "*.py"
▶ Execute this command? [Enter] Run | [N] Cancel

The model translates your request into a command. You approve before it runs.

Example Commands to Try

File Operations:

• "list all python files"
• "find all files named hello"
• "show me the largest files in this directory"
• "count all files in this directory"

System & Network:

• "check if port 8080 is in use"
• "show git status"

Search:

• "grep for hello in this directory"

All commands show a confirmation dialog before running. You can edit the command before executing, or cancel.

Project Structure

LocalCode/
├── main.go              # Go TUI (Bubble Tea)
├── pre-commit.sh        # Pre-commit hook
└── Sources/
    └── afmhelper/       # Swift → Apple FoundationModels
        └── main.swift

• TUI Layer: Go + Bubble Tea
• AI Layer: Apple FoundationModels framework (Swift)
• Command Flow: Model suggests → You approve → Command executes

Development

A pre-commit hook ensures the project builds:

cp pre-commit.sh .git/hooks/pre-commit && chmod +x .git/hooks/pre-commit
git config core.hooksPath .git/hooks

Why does this exist?

Apple's Foundation Models framework is new and under-documented. This project proves it's viable for CLI tools and provides a reference implementation for others building on AFM.

Contributing

POCs have rough edges. Contributions welcome:

1. Fork and create a feature branch
2. Make your changes
3. Open a PR

License

MIT