Show HN: Apfel – The free AI already on your Mac

Apfel, an MIT-licensed Swift 6.3 tool, provides direct access to the ~3B-parameter language model Apple ships with macOS 26 (Tahoe) on all Apple Silicon Macs. Previously accessible only through Siri and system features via Apple's FoundationModels framework, this on-device model now runs locally through Apfel as a CLI tool, OpenAI-compatible server, or interactive chat—with zero token costs, no API keys, and complete privacy since all inference happens on the Neural Engine and GPU.

The tool delivers practical utility with a 4,096-token context window (input and output combined), supports multiple languages (English, German, Spanish, French, Italian, Japanese, Korean, Portuguese, Chinese), and handles real developer needs like JSON output, file attachments, and proper exit codes. Its OpenAI-compatible server runs on localhost:11434, making it a drop-in replacement for cloud APIs, while the CLI supports Unix piping for integration with tools like jq and xargs. Apfel also includes context management with five trimming strategies to work within the model's limited window.

Beyond basic access, Apfel ships with power tools like 'cmd' for natural language to shell command conversion, 'explain' for clarifying code snippets, and 'wtd' for instant project orientation. This transforms Apple's previously walled-garden AI into a versatile, programmable resource for developers who want free, private AI assistance directly in their terminal and applications without relying on external services.