Redde: iPhone/iPad/CarPlay Client for OpenAI-Compatible Models

What changed

The Redde project, developed in Swift, has released version redde-push-1.12.0 on October 31, 2026. This release marks a recent update to the client application designed for iPhone, iPad, and CarPlay. The core functionality of Redde is to act as an interface for AI agents, specifically mentioning compatibility with the Hermes AI agent and any model that adheres to the OpenAI API standard. The application allows for interaction through voice or typed commands, with a key feature being the user's ability to review and approve any actions the AI intends to perform before execution. This approach prioritizes user control and transparency in AI operations.

Why it matters for builders

For developers, Redde provides a foundation for building AI-driven applications on Apple's mobile and in-car platforms. Its support for OpenAI-compatible models means it can leverage a wide range of LLMs. Furthermore, the project highlights support for on-device inference through integrations like llama-cpp and ollama, enabling private and potentially faster AI processing without constant cloud connectivity. The emphasis on a "no account, no tracking" policy is a significant draw for builders focused on user privacy and data security.

Practical impact

Builders can explore Redde to create custom AI assistants for iOS and CarPlay. The client's architecture, which requires explicit user approval for AI-initiated actions, can be a model for developing safe and trustworthy AI integrations. Developers interested in on-device AI can test Redde's compatibility with local LLM setups. The project's Swift and SwiftUI foundation makes it accessible for iOS developers.

Caveats and source limits

The provided repository metadata is limited. Specific details regarding the implementation of llama-cpp or ollama integration, the exact nature of the "Hermes AI agent," or performance benchmarks for on-device inference are not available. The readme_summary indicates "5 AI signals" and "4 developer signals" but does not elaborate on what these entail. The license is listed as "other," which may require further investigation for commercial or distribution use. The project has 101 stars and 8 forks, indicating early-stage community engagement.