Experience with Fellou: The World’s First Agentic Browser

Recently, a new concept called “AI browser” has emerged on the tech scene. Intrigued by the somewhat exaggerated claim that “you no longer need a traditional browser,” I decided to test this new technology and share my experience.

The official name of this tool is Fellou, and you can find the official website at https://fellou.ai/.

On their website, Fellou introduces itself as “The World’s First Agentic Browser.”

It appears that they are preparing for a full-scale service launch, and currently, an invitation code is required to access the platform.

🏷 Key Features of Fellou AI

✅ Website Q&A

Fellou analyzes the content of web pages that users have open and answers questions about them. Examples include webpage summarization, specific information extraction, translation, and more.

✅ Workflow Execution

It automatically performs complex tasks in the browser. Examples include composing emails, creating social media posts, making online purchases, and more.

✅ Deep Search

Fellou searches for and summarizes information on specific topics from across the internet. Examples include researching the latest technology trends, searching for academic papers, and more.

✅ Report Editing

Users can modify existing reports or create new ones. Examples include translating reports into different languages or enhancing content.

✅ Multi-tasking Support

Fellou provides functionality to execute multiple tasks simultaneously.

When I asked Fellou about its capabilities, it confirmed these features. From my direct experience, the primary functions are workflow execution, deep search, and report creation. Let’s examine each of these features in more detail.

🏷 In-depth Feature Analysis

✅ Website Q&A

With Fellou’s Website Q&A feature, you can open a website in a tab and ask questions about it in a side panel. Fellou then analyzes the site to provide summaries and answers to your questions.

While this functionality exists in other AI tools, Fellou’s advantage lies in allowing users to view the website while simultaneously asking questions or requesting analysis. It’s comparable to having an AI assistant embedded in code editors that lets you ask questions while viewing code.

✅ Workflow Execution

This appears to be Fellou’s main feature. I tested it by creating a repository on GitHub.

The process involves configuring tasks step by step and then waiting for execution. When you press “run,” each task is executed sequentially.

Upon execution, Fellou automatically locates GitHub and navigates to the login page. After entering account information and clicking “completed,” it continues with the tasks.

During this process, Fellou automatically analyzes and identifies selectors. It examines the DOM structure of the loaded webpage to automatically determine appropriate selectors.

It then navigates to the creation page and automatically completes the input form. I had requested a repository named “fellou-test-project” set to private status. Since GitHub is a well-known platform, Fellou accurately found the input forms and completed them appropriately.

Finally, it clicks the “create repository” button to generate the repository. I did not intervene at any point in this process.

The repository was created flawlessly on the first attempt, which was somewhat surprising.

The process took approximately 2–3 minutes, likely due to the time needed for analysis and task processing.

✅ Deep Search & Report Creation

When performing a deep search, Fellou simultaneously opens multiple subwindows, extracting or summarizing information from each. It collects and processes information from multiple sources simultaneously, typically compiling this information into a report.

For report creation, Fellou generates actual code to construct a webpage for browser display.

The reports produced are remarkably detailed and comprehensive — far more extensive than what typical AI tools could generate given token limits. The content is thorough and high-quality.

Examples of generated reports include:

• GUI Agents: A Comprehensive Research Summary: https://chat.fellou.ai/report/d871e5ef-7909-4874-91a2-58a04ec59e18
• Fellou AI Research: https://chat.fellou.ai/report/040c2583-bc5a-4ab8-89bd-160ff5c8c3f5 (Although I only asked for recent trends, the content wasn’t entirely satisfactory.)

🏷 Concluding Thoughts

I tested Fellou out of curiosity about claims like “a new era of browsers has arrived” and “this is an AI browser” — wondering if such capabilities were genuinely possible.

My personal impression is that Fellou appears to be a combination of AI + MCP + Playwright + Webview. While the actual implementation is likely more complex, it feels like an integration of various distinct technologies.

Currently, users typically need to switch between different tools to complete complex tasks, but Fellou handles everything in one place, which could be quite convenient once users adapt to it.

A notable advantage is Fellou’s ability to generate extensive content without the token limitations typical of other AI tools.
Though this may be because it’s not yet a full production service, the output is impressively comprehensive. (I was genuinely surprised by the volume of results initially.) However, this does mean that generating results takes time — approximately 5–10 minutes for report creation.

Fellou also handles basic Q&A and simple coding tasks well, though it seems somewhat slower than alternatives.

For a pre-release service, the functionality is quite impressive. With additional features and improved stability, it could become a powerful “AI browser” in the future.

That’s all for today. Have a great day!

This essay was originally published on Medium and migrated to KanApp Blog.