People use AI for increasingly ordinary material: a contract draft, study notes, a CSV export, a long report or a folder of research. The value is obvious, but so is the question that follows: where did that document go?
Local-first software does not make every privacy problem disappear. It does, however, make the main data path easier to understand. When document retrieval and core processing happen on the Mac, a useful task does not automatically become a document upload.
Privacy starts with a visible boundary
A good local workflow tells the user what remains on the device, what requires a network connection and what optional components are being downloaded. It should not hide a multi-gigabyte model inside an innocent-looking update or silently switch to a remote service.
Neravia keeps imported documents on the device and does not require an account for its core document, writing, extraction and translation workflow.
That distinction matters for consultants, students, researchers and small teams who often work with material that is not public but does not justify a complex enterprise AI deployment.
Source-aware answers are more useful than confident answers
When an assistant answers a question about a document, fluency is not enough. The user should be able to inspect the passages that shaped the answer. Neravia retrieves relevant sections locally and surfaces the source material so the result can be checked rather than simply trusted.
This is especially useful for long PDFs, structured exports and notes where the real job is not creative writing. It is locating a fact, understanding context and turning the result into a next action.
Hardware limits should be honest
Large local models need memory, storage and power. Neravia treats those limits as product information, not fine print. Its optional Bonsai 27B MLX 1-bit model is approximately 5.16 GB and is enabled only on devices with at least 10 GB of physical memory. The app also considers temperature and Low Power Mode before advanced local inference.
A smaller Mac can still use the core workspace. The point is not to pretend every device can run every model. The point is to give each device a useful, understandable experience.
The Mac is a natural home for this work
A Mac already holds the documents, drafts and research that people want to understand. A native workspace reduces the friction between selecting a file and doing something useful with it. It also gives local AI access to the machine resources it needs while keeping controls close to the user.
Neravia is our attempt to make that relationship calm: import a document, choose a task, review the result and keep control over the model and the data path.
Try the current release
Neravia AI is available on the Mac App Store. The free version includes the core workspace with a small non-personalized advertising banner. Neravia Pro removes ads for $0.99 per month. Availability and local currency pricing are shown by Apple for your storefront.