Free and open source Nanobrowser, has become an "invisible master" of Web automation?

What is Nanobrowser

As a test engineer, I am often tortured by tedious web operations in my daily work. For example, repeatedly switching between different pages, mechanically filling out various forms, and filtering information like looking for a needle in a haystack... These seemingly insignificant tasks consume a lot of my time and energy, and my efficiency plummets. It is also easy to make human errors due to fatigue. I have been looking for a powerful tool that can help me get rid of these repetitive tasks, until I foundNanobrowser.

NanobrowserIt is an open sourceChromeExpanded tools, focusing onAIThe web page automation driven by the browser is like hiring an intelligent assistant to complete various complex web page tasks independently. It is completely free, all operations are run in the local browser, and there is no need to worry about the leakage of private data. It also supportsDeepSeek,OpenAI,Anthropic,GeminiYou can flexibly configure the mainstream large-scale language models according to your needs and budget.APIKey, choose the most suitable model.

Functional test: Is it really as magical as it is said to be?

It is not enough to just listen to the introduction, I ( TesterRoad ) decided to test it myself and see if it works.NanobrowserIs it really as magical as the legend says?

Multi-agent system: clear division of labor and tacit cooperation

Open and installNanobrowserExtendedChromeThe browser, a simple sidebar interface appears in front of me, which is the main position for interacting with the intelligent agent. I decided to give it a complex task: on the e-commerce platform, help me choose a laptop with a price between 5,000 and 8,000 yuan, a memory of 16GB or more, a hard disk of 512GB or more, and a user rating of 4.5 points or more, and sort out the three most cost-effective ones.

After entering the command in the sidebar, the plannerPlannerIt quickly starts analyzing the task and developing a detailed execution strategy. It first determines which e-commerce platforms need to be visited, and then plans the search steps and filter conditions.NavigatorJust followPlannerThe formulated strategy automatically opens the pages of major e-commerce platforms in the browser, accurately enters the search keywords, clicks the filter button, and fills in the filter conditions such as price, configuration, and rating. In this process, every step of the operation can be clearly seen in the browser, just like an invisible hand helping me click.

Finally, the validatorValidatorCome on stage and check carefullyNavigatorThe product information obtained ensures that the data is accurate and no important details are missed. It will recheck key information such as price range, configuration parameters, user ratings, etc., and eliminate products that do not meet the requirements.NanobrowserThe information of the three most cost-effective laptops was quickly sorted out and presented to me. The whole process was completed in less than two minutes. If I had to screen them manually, it would probably take me more than half an hour.

Interactive sidebar: natural conversation, real-time feedback

During the use, I was deeply impressed by the interactive sidebar. Its interface design is very simple and intuitive, just like the chat software we usually use. Entering commands and viewing replies are clear at a glance. When I put forward task requirements to it, it can quickly understand my intentions and give accurate responses. Moreover, during the execution of the task, it will update the task status in real time, so that I can keep abreast of the progress.

Once, I asked it to help me collect news about artificial intelligence in the past week and sort out the key points. As soon as I entered the command, the sidebar showed "Task received, strategy in progress", and then immediately displayed real-time status such as "Accessing news websites", "Screening related articles", and "Extracting key points". If there is a problem, it will also give feedback in time. For example, if a website cannot be accessed, it will prompt me with possible reasons and try to find other alternatives. This real-time interactive experience makes me feel that I am not dealing with a cold tool, but having an intelligent assistant by my side, communicating at any time.

Task automation: free your hands and increase efficiency

To test the ability of task automation, I gaveNanobrowserAssigned a repetitive and tedious task: extract the names and contact information of all customers over 30 years old and living in a specific city from a web form containing a large amount of customer information, and organize them intoExcelIn the past, I had to manually check the table line by line, filter out the customers who meet the criteria, and then copy and paste them intoExcelIt is not only time-consuming and laborious, but also prone to errors.

But withNanobrowser, everything becomes easy and simple. I just need to enter the command in the sidebar, and it will start working immediately. It automatically recognizes the web form, quickly and accurately extracts the relevant customer information according to the filter conditions I set, and then automatically opensExcelThe software filled in the data into the form and saved the file for me. The whole process was fully automated. I only needed to sit in front of the computer and watch it complete the task in an orderly manner. In just a few minutes, I completed the work that would have taken one or two hours. The efficiency was improved several times. Moreover, since it was machine-operated, human errors were avoided and the accuracy of the data was guaranteed.

Multiple LLM support: flexible choice, each with its own strengths

NanobrowserSupports a variety of mainstream large-scale language models, which allows me to flexibly choose the most suitable model according to different task requirements and cost considerations. I usedOpenAIofGPT-4andAnthropicofClaudemodels to perform the same task and compare their performance.

When dealing with tasks that require creativity and deep understanding, such as writing an article about technology trends,GPT-4It has demonstrated strong language generation capabilities, with articles that are logically clear, rich in content, and fluent in language expression. When dealing with tasks that require high accuracy and facts, such as compiling financial data reports,ClaudeIt performs very well, accurately analyzing data, giving accurate conclusions and suggestions, and paying close attention to details.LLMSupported by the way, I can give full play to the advantages of different models, soNanobrowserPerform better in various tasks.

Detailed tutorial: Even a novice can get started in seconds

Can't wait to experience it yourself?NanobrowserDon't worry, I will teach you how to install and use it step by step, even a technical novice can easily get started.

Installation Steps

1. 1.  Download the file : Open your browser and visitNanobrowserOfficialGitHubWarehouse, find the latest version on the pagenanobrowser.zipfile, click Download.

https://github.com/nanobrowser/nanobrowser/releases

1. 2.  Unzip the file : After downloading is complete, find the downloadednanobrowser.zipfile and unzip it.
2. 3.  Enable developer mode : Turn onChromeBrowser, go to the extension page. Turn on "Developer Mode".
3. 4.  Load extension : In the upper left corner of the extension page, click the "Load unzipped extension" button. In the pop-up file selection window, find and select the file you just unzipped.nanobrowserfolder, and then click Select Folder.

Of course, you can also go directly to the Chrome App Store to search and download

Configuring the Agent Model

After the installation is complete, we also need to configure the agent model so thatNanobrowserAbility to use a large language model of your choice.

1. 1.  Open the sidebar : ClickChromeIn the browser toolbarNanobrowserIcon, openNanobrowser's sidebar.
2. 2.  Add API key : In the upper right corner of the sidebar, click the settings icon, and in the pop-up settings menu, findAPI Keysoption, click it. Then, enter the correspondingAPIFor example, if you useOpenAIModel, just inputOpenAI APIkey; if usingAnthropicofClaudeModel, just enterAnthropic APIKey.
3. 3.  Select model : AddAPIAfter the key,Model SelectionNext, findModeloptions. Here you can setPlanner,Navigator,Validator) to select the model to use. For example, you canPlannerchooseGPT-4Model to obtain more powerful mission planning capabilities;Navigatorchoosegpt-4oModel, making it more accurate when performing web page operations.

Command input and task execution

Everything is ready, now you canNanobrowserIt's time to show off your skills! In the input box on the sidebar, enter your task instructions, just like chatting with a friend, and describe the task you want to complete in natural language. Below TesterRoad gives you some common task instruction examples:

1. 1.  Data capture : "Extract movie names, ratings, and director information from Douban Movie Top 250 page and save as CSV file."
2. 2.  Information monitoring : "Monitor the latest notes about 'new beauty products' on Xiaohongshu and send them to me once every hour."
3. 3.  E-commerce shopping : "Search for 'sports shoes' on Taobao, filter out the top 10 products with a price range of RMB 300 - 800, and compare their price/performance ratio."
4. 4.  Office Automation : "Log in to my corporate mailbox, mark the emails with the subject "Important Meeting" as read, and forward them to my colleagues."

After entering the command, click the Send button.NanobrowserIt will start to analyze your instructions, formulate a task strategy, and automatically execute the task in the browser. During the task execution, you can view the task status in real time in the sidebar to understand the progress of the task. If the task is executed successfully,NanobrowserThe results will be displayed in the sidebar; if there is a problem, it will also provide timely feedback, tell you the possible cause, and try to find a solution.

Comparison with similar tools: What are the advantages?

There are many web automation tools on the market, such as the well-knownOpenAI Operator,Browser-Use ,NanobrowserCompared with them, the advantages are quite obvious.

andOpenAI Operatorcompared to,NanobrowserThe biggest advantage is that it is free and open source.OpenAI OperatorThe monthly subscription fee is not low, which is a considerable expense for individual users and small teams.NanobrowserIt’s completely free, there is no subscription fee, just install the extension, configure yourLLM APIKey, you can use the top web automation features at a much lower cost, so that everyone can easily enjoyAIIn terms of privacy,OpenAI OperatorIt runs in the cloud and the data is stored in the cloud server, which poses a risk of privacy leakage.NanobrowserAll operations are performed in the local browser.APIBoth keys and operation data are stored locally, so there is no need to worry about sensitive information being leaked. This is great news for privacy-sensitive users.

and based onPythonDeveloped byBrowser-Usecompared to,NanobrowserIt is superior in terms of ease of operation.Browser-UseAs aLLMIntelligent browser automation tool library requires certainPythonIt requires basic programming skills, and for ordinary users, the threshold is high.NanobrowserIt provides an intuitive interactive sidebar, and users can use natural language commands to let it complete various web page tasks, just like chatting with an intelligent assistant. There is no need to write any code, and it is truly zero-threshold to get started.Browser-UseIt can realize functions such as web page automation, multi-label management and visual recognition, butNanobrowserThe multi-agent system is more intelligent and flexible. Different agents work together to better handle complex web workflows. When encountering obstacles, it can also automatically adjust strategies to ensure smooth completion of tasks.

Summary and Outlook

After this period of in-depth use and testing,NanobrowserIt brings me a real leap in efficiency. It is like a capable assistant who understands my mind and helps me easily complete those tedious and repetitive web page operations. Whether it is the tacit collaboration of the multi-agent system, the convenient communication of the interactive sidebar, or the freedom of hands brought by task automation, as well as multipleLLMFlexible options supported, every function makes me feelAIThe power of driving web automation.

From the perspective of software testing,NanobrowserThe stability and accuracy of the tool are excellent. In multiple tests, it can accurately perform tasks according to the set instructions, with few errors or abnormalities. Moreover, its open source and free features, as well as its high attention to privacy, make it stand out among many similar tools and become the only choice for users who value efficiency and privacy.

As far as I know,NanobrowserThe development team of is working hard to develop new features. In the future, it will integrate more large language model providers to give us more choices; the addition of a security firewall will make browsing operations safer and more reliable; memory optimization is also in the plan, which can reduce the use of tokens while maintaining excellent context awareness; the session playback function allows us to review historical tasks efficiently and reduce token consumption; dedicated agents for complex domain tasks are also under development, which I believe will bring great convenience to work in specific fields.

If you are like me, and are bothered by cumbersome web page operations, you might as well tryNanobrowserI firmly believe that it will bring unexpected surprises to your work and life, allowing you to say goodbye to repetitive work and embrace efficient and convenient digital life. Go and experience it now, and remember to share your experience in the comment area!