A new AI assistant called Local Code has launched for Mac, enabling users to run powerful AI models entirely on their machines without needing to write any code. Created by the same team behind Draw Things, this free, open-source app allows macOS users to download models and process tasks locally rather than relying on cloud infrastructure.
What Local Code Can Do
Local Code supports a variety of use cases: generating presentations, creating images and video content, analyzing spreadsheets, and working with Word and PowerPoint files. It also has built-in browsing capability so the assistant can navigate websites to carry out online tasks. With user approval, the app can access data from Mail, Messages, and Notes, as well as modify Calendar and Photos. Its access is limited by Apple’s App Sandbox to protect system integrity and privacy.
Models, Sizes, and Desktop Use
The app includes two main model families: Qwen and DeepSeek. Models scale based on your Mac’s memory and disk space. For example, Qwen 3.8’s 27-billion parameter version needs at least 16GB of RAM, and uses between roughly 10GB and 21GB of storage. DeepSeek models are considerably larger, with certain versions requiring up to 89GB or even 376GB for download.
Pricing, Cloud Options, and Availability
Using Local Code locally is free. For those who prefer or sometimes need cloud access, the app supports external API keys or prepaid credit plans starting at $9.99. When operating in cloud mode, the prompts and relevant files are sent to the external provider, unlike the fully local model which keeps everything on the Mac.
The app emerged from a public beta that began on September 12, and a direct download for Mac is already available. A Mac App Store version is expected in the near term.
Local Code joins a growing wave of AI tools that put privacy and user control first, running inference on local machines and reducing reliance on cloud services. For Mac users worried about data security or ongoing costs, this is an especially attractive option.
Why it matters: As AI becomes central to user workflows, many demand that their tools respect privacy by keeping data on device. Local Code meets that demand while still offering flexibility—local or cloud-based operation depending on resources and preferences. What to watch for: the maturity of its performance on Macs with limited memory, how DeepSeek models run in practice, and how seamless the upcoming App Store version will be.
