Napcar

Local AI for the web.

A Chromium-based browser that lets websites run language-model features on your Mac through local models instead of cloud APIs.

Keep one call site: the SDK uses Napcar's local model when it's available, and falls back to a cloud provider you choose in any other browser. See how the fallback works.

Download for macOS Read the JavaScript docs

Local model execution

Websites run language-model features on your machine — text never has to leave your Mac.

Ollama-powered backend

A bundled Ollama-compatible service manages models. No separate install, no terminal.

Browser permissions for AI

Local AI is a browser capability gated per-site, like camera, microphone, and geolocation.

OpenAI-compatible endpoint

Point existing clients at https://local.napcar.ai/v1 — synthesized in the browser, never the network.

One call site, every browser

The SDK runs the local model in Napcar and falls back to a cloud provider you choose (OpenAI by default, or any OpenAI-compatible API) everywhere else.

Model rankings at chrome://llm

Install, rank, and keep models warm by task: chat, code, summarization, embeddings, vision.

Built for privacy

No prompt or response telemetry by default. A local, metadata-only audit log you control.