GitHub 5k+ Star open source project | AI-driven browser automation tool Nanobrowser

Written by
Caleb Hayes
Updated on:July-02nd-2025
Recommendation

Nanobrowser, an AI-driven browser automation tool, makes web page operation automation a reality.

Core content:
1. AI agents work together to dynamically adjust task strategies
2. Local operation ensures privacy and security, and supports multi-model optimization of cost and performance
3. Quick start guide, full process guidance from download and installation to task execution

Yang Fangxian
Founder of 53AI/Most Valuable Expert of Tencent Cloud (TVP)

Nanobrowser's core features

  • •  Multi-agent collaboration : Through the collaboration of three AI agents, Planner, Navigator, and Validator, task strategies can be adjusted dynamically like real people.
  • •  Privacy and security : All operations are run on the local browser, and data does not need to be uploaded to the cloud, completely eliminating the risk of leakage.
  • •  ? Multi-model support : Compatible with mainstream large models such as OpenAI, Anthropic, Gemini, DeepSeek, etc. Users can freely assign different models to optimize cost and performance.
  • •  ? Intelligent error correction : When encountering unexpected situations such as verification codes and page changes, the strategy is automatically adjusted, and the task success rate is as high as over 90%.
  • •  ? Completely free and open source : Based on the AGPL-3.0 protocol, the code is transparent and traceable, and it is the best alternative to OpenAI Operator.


Why you need Nanobrowser

  • •  Operators : Automatically generate competitor reports and capture social media data, saving you from staying up late to copy and paste.
  • •  Developers : Crawl GitHub trending projects and automate testing of web page interactions with one click.
  • •  E-commerce parties : monitor commodity price fluctuations in real time and automatically grab limited-time discounts.

5-minute quick start guide

  1. 1.  Download and install :
  • • Download the latest version of the extension pack from GitHub  https://github.com/nanobrowser/nanobrowser/releases/tag/v0.1.4
  • • Chrome browser opens chrome://extensions/, enable developer mode and load the unpacked extension folder.


    • 2.  Configure the model key :
      • • Click the toolbar icon to open the sidebar, add the DeepSeek API key in the settings, and assign models to different agents.

    • 3.  Execute the task :
      • • Enter natural language instructions (such as “grab today’s popular Python projects on GitHub”), and the AI ​​agent will automatically parse and execute the complete process.


    • 4.  Monitoring and adjustment :
      • • View task status in real time through the interactive sidebar and make adjustment requests at any time (such as "exclude projects with less than 1000 stars").


      Precautions

      • •  Hardware requirements : A stable network environment is required. For some complex tasks, it is recommended to use high-performance models such as GPT-4.
      • •  First time use : You need to authorize the browser to record, but all data will only be processed locally.
      • •  Dynamic adjustment : If a task fails, you can intervene manually or let Planner automatically retry a new strategy.

      Open Source and Future Development

      Nanobrowser is completely open source, and developers can contribute code or customize exclusive features. Its team is planning to launch:

      • •  Expanded model support : Integrate more large models such as Claude, Zhipu AI, etc.
      • •  Security firewall : prevents malicious script injection and ensures the safety of automated operations.
      • •  Session replay function : efficiently review historical tasks and optimize token consumption.

      Project address : https://github.com/nanobrowser/nanobrowser