ChatLLM Web
ChatLLM Web is a browser-based, private local model studio that runs chat and agent workflows entirely in the browser using WebGPU and WebLLM, with no backend, account or telemetry.
ChatLLM Web combines a chat interface, a local agent workspace and a local model catalog in one browser-native application. It inspects the current browser, recommends a compatible model, and runs WebLLM in a dedicated worker so inference happens on-device; conversations, selected files, tool results and generated artifacts stay on the device.
Each conversation can switch between chat and agent modes; in agent mode the app runs native WebLLM function calling with three Hermes models and lets the model inspect selected files, search context, do calculations and work with saved artifacts through sandboxed local tools, with every artifact write shown as a line diff before it is saved. The catalog offers 20 curated models to start, with an advanced view exposing 65 logical models backed by over 160 official WebLLM records, and device capabilities pick a compatible quantization automatically.
The application is static, served by Cloudflare Pages, with model files fetched directly from Hugging Face and WebLLM's own library; it can also be installed as a progressive web app.