Nanobrowser setup with DeepSeek: browser automation

The original video is in Chinese. This tutorial covers Nanobrowser, an open source browser extension that the presenter describes as a free alternative to OpenAI Operator. It walks through installation from the Chrome Web Store, pinning the extension and adding a model provider. The presenter states that Chrome and Edge are supported; functionality in other browsers is not guaranteed.

The demonstrated setup uses a DeepSeek API key and selects DeepSeek-chat for both the planner and navigator. The planner breaks a request into steps, while the navigator operates the browser. A short explanation of temperature follows, though the demo keeps the default settings. Ollama appears among the provider options, and the presenter mentions a local DeepSeek deployment without showing that setup here.

The first task asks the AI agent to collect article titles and links from demo2easy.com. The reported output contains 13 articles across two pages, which the presenter copies into a text editor. Some API errors trigger retries. The presenter reports a runtime of more than ten minutes and suspects network conditions contributed to the delay. The second task searches for 2025 Chinese electric vehicle sales data and the top three brands.

Although the introduction claims data stays local, the walkthrough describes sending tasks and selected page information to the configured AI provider. This API-based demonstration does not establish an entirely private or offline workflow. The presenter also notes that model calls may cost money and recommends checking the output manually.