Supercharge Your Browser with Mind Studio's AI Agents

This article is about a Chrome extension from Mind Studio that literally gives your browser superpowers. They've built a Chrome extension that offers these AI agents, letting you do so much more with your web browser. There are curated lists of agents, and you can even build your own. This article will show you how to install it, demo some AI agents, and walk you through how to build one.

Installation

Let's proceed with the installation.

  1. Navigate to the main website page.
  2. Click on the installation button, which will take you to the Chrome Web Store.
  3. Add the extension to your browser. It works on all Chromium-based browsers, including Arc.
  4. Once the extension is added, pin it. Clicking on it will open up the side menu.
  5. This menu gives you access to all the agents and allows them to interact with the browser.
  6. Next, log in. I'm logging in with my Google account.

You can clearly see that they've pre-listed everything. You can open the doc icon, run the agents, and pin your favorites.

Available Agents

Let's look at some of the agents available here, and it's a pretty crazy selection.

  • Research Agents: Including one for deep research. You'll notice some agents are free (they don't have anything written underneath them), while the paid ones are clearly labeled. There are some really cool agents as well, like a fact-check agent and a product comparison agent.
  • Content Creation Agents: For example, there's a TLDDR agent that instantly summarizes your documents and articles. You could do the same thing by copying everything into chat GPT, but this way it feels a lot more seamless.
  • Social Media Agents: These can take content directly from a page and transform it into posts for LinkedIn, Instagram, Twitter, or even turn a YouTube video into an article.
  • Image Generation Agents: The fun thing is they're also free.
  • Developer Agents: Though I didn't find them all that useful, the API cheat sheet agent does look promising. It takes any API documentation and generates a cheat sheet explaining how to use it. It's helpful if you want to understand the code yourself, but with tools like cursor and windsurf, the AI can just implement the code for you, so there's not really a need anymore.

There are also some paid AI models and other fun tools to explore.

Agent Demos

Now that we've installed it and seen some examples, let's run one on a YouTube video. We've got our YouTube publication open, so let's click on one of our articles in the for creators section. If you scroll ahead, there's a YouTube to X thread agent; that's what we're going to run on this article.

To run the agent, just click on "run." It'll open up this interface. It's actually a pretty nice UI. The agent starts running and continues processing on its own.

A cool thing that just popped up is the notification feature. It actually lets you know when the agent is done. Some agents take a while, like 2 to 3 minutes, so instead of waiting around, you'll get a notification once it finishes. Here we can see that the thread has been generated. This really makes things easier for me. Usually, when I have to post my articles on X, I either write the thread manually or generate it from my script. With this, it's a seamless process. I can run the agent on my article, copy the results, and post them to X.

This isn't the only tool. You can see it pulled the article's URL, so it's likely extracting the transcript directly. There are tools online that do the same from a YouTube URL, and it probably uses something similar to generate the thread. This is just one of many tools. It also works for LinkedIn. There's even a YouTube video description generator that includes chapters and timestamps, which is super helpful.

Here's another cool AI agent you can use. I've opened up this MacBook Pro specifications page. It's the actual product page, and it has all the specs listed. That helps the AI agent, which is designed to find competitors for your product. Let's run it and see what competitors it suggests for the MacBook Pro.

It's done, and it actually pulled all the features and gave us a competitor analysis. It highlights the strengths and weaknesses of the MacBook Pro compared to competitors. This is actually pretty good and useful. You get accurate competitor suggestions based on the specs it reads directly from the web page.

Building Your Own Agent

Now let's move on to the building part. I made this agent called "Word Query." It's something I wanted for myself. Say you're browsing and come across a word you want to learn more about. Normally, you'd copy it, go to chat GPT, and ask. With this tool, you can simply select the word, like I've selected "vibe coding," and run the tool.

Let's wait while the agent runs. It prompts you to type in a question. If you enter one, it applies that question to the selected word. If you leave it blank, it gives you the meaning. Since we selected "vibe coding" and didn't ask anything, it returned the meaning of the term.

Steps to Build the Agent:

  1. Go back and open the menu, then go to the "Build" tab.
  2. You'll see "Published" and "Drafts." Expand it and find the "Create a new agent" section.
  3. Clicking that opens a new project in a visual builder, like building automations.
  4. Inside the builder, you'll see a "Start" and "End" block. Click on the "Start" block to configure triggers and run mode.
  5. Select "Browser Extension," and this will give you useful launch variables like URL and user selection. These let you grab data directly from the web page.
  6. Next, add a "User Input" block. Create a new input with the variable name user_question and label it as "Question."
  7. Now, add a "Generate Text" block. You'll see other options like "Generate Image," which lets you create images from selected text.
  8. Select "Generate Text," and in the prompt area, use double curly brackets to insert variables. First, reference the user_selection.

  9. Then, write this conditional prompt:

    if the  is empty, return the meaning of the . if not, return what the  asks.

  10. This prompt is sent to the model. You can also select which model to use. The free ones are Claude 3.5 Haiku and GPT-4o mini. For now, I'll leave it as it is.

  11. Under "Output Behavior," you should choose to display the result.

Note: There was a bug. Even though behavior was set to "End Session," the agent still returned a JSON output. As a temporary fix, I removed the "End" block. It didn't officially end the agent, but I could go back, so it worked. I'll look into a more permanent fix.

Once you're done, click "Publish" to make your agent functional.

Whether you're just exploring AI agents for the first time or already deep into the world of automation, I think this extension has a lot to offer.