Nanobrowser
An open-source browser AI agent for Chrome and Edge that automates web tasks using local models through Ollama or cloud providers with your own API keys.
Nanobrowser is an open-source AI agent that automates web tasks inside Chrome or Edge. It's for people who want to delegate repetitive browsing or research while choosing which models handle the work. The extension runs in your browser and is an alternative to OpenAI Operator, with an Apache 2.0 license.
Model choice is central to its design. You can run models locally through Ollama or a custom OpenAI-compatible provider, keeping model processing on your machine. Cloud options include OpenAI, Anthropic, Gemini, Groq and Cerebras, using your own API keys. The browser automation runs locally in either case; cloud models handle inference through the selected provider.
Separate agents plan tasks and carry out browser actions. The Planner can revise its approach when it encounters obstacles and direct the Navigator to try another route. You can assign different models to these roles. Local model recommendations include Qwen3-30B-A3B-Instruct-2507, Falcon3 10B, Qwen 2.5 Coder 14B and Mistral Small 24B.
A chat side panel shows task progress and lets you ask follow-up questions about completed work. Conversation history keeps earlier interactions accessible. Tasks include collecting news headlines, researching GitHub repositories and comparing shopping options across websites. Chrome and Edge have full support; Firefox and Safari aren't supported.
