Skip to content

GlossaryDefinition

What is local-first AI?

Local-first AI runs models and stores data on the user's own device by default, using cloud services only when the user opts in.

The term comes from local-first software, where the primary copy of your data lives on your device and the network is optional. Applied to AI, it means embeddings are computed locally, indexes are stored locally, and memory is written to local files. The application works with the network off. Cloud models or stores can be added, but adding them is a choice with a visible consequence: the data you send goes to that provider.

Local-first has costs and benefits. Costs include smaller models and a machine that has to do the work. Benefits include no per-token charge for the bundled local embedding workflow, no required project-document upload for local preparation and search, and less dependence on a hosted memory service staying in business. License activation still contacts the licensing service, connected AI tools receive the context you request, and optional cloud connections receive selected data when enabled. Rewind shut down in December 2025. Limitless was acquired by Meta. Khoj Cloud sunset in April 2026. A folder on your disk remains under your control even when a third-party service changes.

How Central Brain handles local-first ai

Central Brain is local-first by default: bundled embedding model, local JSONL store and local memory files, so preparing files, indexing, search and memory run air-gapped, with no network connection. The license is re-checked online every 14 days (Team and Enterprise can get a build with no online check); optional cloud connections and cloud-hosted AI clients need internet.