If you build iOS apps with AI, you've probably hit a wall the second you leave your text editor. Writing logic is fast. Getting the agent to manage simulators, configure a local database, and push a build to App Store Connect is a nightmare of brittle shell scripts.
Blitz fixes that deployment gap. It is a native macOS app built specifically to hand the iOS lifecycle over to your agents. You can grab the source straight from their GitHub repository.
The app requires macOS 14 or later and wires directly into Xcode. The real magic comes from its built-in Model Context Protocol server. You boot up Blitz, and clients like Claude Code instantly know how to build, test, and package your code.
It strips back the awful deployment process. You don't have to teach an LLM the dark arts of code signing. You let the agent talk to the Blitz API, and the app handles the heavy lifting right on your local machine.
I'm a big fan of their approach to security. They've sanded down the telemetry to absolute zero. The app makes no tracking calls and has absolutely no phone-home logic.
The MCP server binds tightly to localhost, so it's completely invisible to your broader network. It also stays out of your personal data. It doesn't poke around your contacts or location. (Screen capture is completely locked to the iOS Simulator window.)
They even let you audit the CI pipeline. Every release binary comes from a public GitHub Actions verification flow and the public build workflow, with published checksums. If you want to be completely certain about what you're running, you can pull the source, build it yourself, and compare the hashes.
I think this represents exactly where agentic coding needs to go next. We have plenty of text generators. We need more bridges that let agents safely operate the physical machinery of shipping software.