Overview
FMProxy is a local OpenAI-compatible chat completions proxy for Apple Foundation Models on macOS 26 Tahoe. It requires macOS 26.0 or later, Xcode with the macOS 26 SDK, and Apple Foundation Models available on the Mac. The project can be built with xcodebuild or the provided build.sh script, and the standalone Apple Silicon binary can be started from the repository root. By default it listens on 127.0.0.1:8080, with command-line options for host, port, logging suppression, structured-output validation, and help. The server exposes GET /health, GET /v1/models, and POST /v1/chat/completions. The chat endpoint accepts system, user, and assistant roles; stream: false returns regular JSON, while stream: true sends OpenAI-compatible Server-Sent Events followed by data: [DONE]. Token statistics are returned as 0. Generations are processed one at a time in FIFO order, while health and model-list requests remain available. Structured output supports response_format values of json_object and json_schema, with optional full JSON Schema validation disabled by default. HTTP limits include headers up to 64 KiB, bodies up to 10 MiB, and a 30-second timeout for incomplete requests.
Pricing
Free
Detailed plans have not been confirmed in our catalog. Check the official website for current limits and billing terms.
Visit WebsitePrices and limits may change. Confirm the currency, billing period, seat minimum and usage caps on the official website.
Use Cases
- Developers use FMProxy to expose Apple Foundation Models through an OpenAI-compatible chat completions API.
- Teams use FMProxy to run a local proxy on 127.0.0.1:8080 with configurable host and port.
- Applications use FMProxy to stream chat completions via OpenAI-compatible Server-Sent Events.
- Developers use FMProxy to request structured JSON output with response_format json_object or json_schema.
- Teams use FMProxy to validate generated JSON against a requested schema by enabling --validate-structured-output.
- Operators use FMProxy to monitor completed requests through compact console logs without logging request bodies or model content.
Features
AI assistant
Natural language interface
Workflow automation
Reviews
No reviews yet. Be the first to share your experience!