Forge v1.0.0 - source-installable public launch

Forge v1.0.0 is the public source-installable launch.

Forge is a native macOS workbench for Apple Silicon local LLMs. It runs MLX and in-process GGUF models, scans model folders from disk, manages prompt libraries and presets, dispatches agents, speaks MCP, and can expose a local OpenAI-compatible API for tools.

Install from source:

git clone https://github.com/christopherbattlefrontlegal/swift-mlx-forge.git
cd swift-mlx-forge
./scripts/install-forge-app.sh

Included in this launch:

• Native SwiftUI app for local MLX and GGUF model work.
• Source installer that builds, bundles, ad-hoc signs locally, installs /Applications/Forge.app, and opens it.
• Multi-model slots, Room split view, and coordinated multi-model rounds.
• Prompt folders, Smart Prompt Select, system-prompt presets, design prompt builder, Brave research, agent dispatch, OpenRouter coding loop, and headless command composer.
• Local OpenAI-compatible API server and MCP client path.
• Public landing page and launch kit.
• Developer security check script and CI toolchain gate.

Boundary:

This release is source-installable. It intentionally does not attach a public drag-and-drop .app binary yet. A public binary should be signed with a Developer ID Application certificate and notarized before being attached to a GitHub release.

Validation performed before release:

• ./scripts/security-check.sh --developer
• swift build --product mlx-forge
• GitHub Pages deploy of the public landing page
• Cloudflare Pages deploy to the new mlx-forge Pages project

MLX-Forge

MLX Forge is for Mac users who keep models on disk and want local inference to feel native. It scans MLX and GGUF model folders, lets you load and unload models deliberately, exposes prompt and sampling controls, stores optional cloud provider keys in Keychain, and can serve a local OpenAI-compatible endpoint for agent tools.