đ§Ș Table Studio is in beta, and it is on for every account. We are actively improving this feature daily.
Table Studio is the new way to build a robot in Browse AI. Point an AI agent at a URL and it reads the page, builds the table structure for you, and fills it with sample data. You review, edit, and save. From then on the robot extracts that data on demand or on a schedule, like any other Browse AI robot.
đĄ You do not need to know exactly what you want before you start. The agent proposes the structure from what is on the page.
Watch & learn
Watch the full flow, from a URL to a saved robot:
Is Table Studio the right tool for your dataset?
Currently, Table Studio works best when the data you want is visible on the page as soon as it loads, like a product listing, a directory, a search results page, or an article.
â Start in Robot Studio instead when the robot has to interact with the site to reach the data, including:
clicking through pages
filling inputs or search forms
navigating (i.e. dynamic content)
if the data you need is behind a login and requires you to enter a username and a password (i.e. cannot be access via cookies)
That is what Robot Studio's visual point-and-click training is built for.
đĄ If the data you are extracting is behind a login and you are already logged in on Chrome, check This website needs logging in, choose Login with my session cookies, then select Table Studio.
Not sure? Try Table Studio first. If it hits something it cannot handle, it tells you why and routes you to Robot Studio.
How to train an agent using Table Studio
Step 1: Point it at any URL
From your Browse AI dashboard, click Build New Robot.
Enter the Origin URL: the page that shows the data you want.
Click Start Training Robot, then choose Table Studio.
đĄ Pick the right starting URL. The closer your Origin URL is to the page that shows the data, the better the result. Point it at example.com/products, not the homepage.
Step 2: The AI agent builds your table
The agent takes it from there. It loads the live page, reads its structure, works out which values appear once and which repeat as lists, proposes the tables and columns, and extracts sample rows.
â±ïž This can take 5 to 10 minutes depending on the page. You will see progress at the bottom of the screen while it works. There is nothing to configure at this stage; let it finish.
Step 3: Review and edit the data structure
Once the agent is done, you can review, delete, add or edit the data it has captured.
đĄ If the table includes a list, youâll be asked How many items do you want from [list] at most? Pick 50, 500, 5,000, or type a number. Then you can save.
How to add data
Add a column to add additional data using plain language. You can add:
Single text
A list
A screenshot
Select the data type you want to capture. Name it and add Instructions in plain language, for example âthe listed price before any discountâ. Then click Preview. The agent uses your description to find the right element on the page.
How to remove data
The agent by default will include all data on the page it thinks is relevant to you. To remove columns you don't want to capture, select the column and click Delete column.
Step 4: Save to unlock monitoring, bulk actions, and more.
When it looks right, click Save Robot in the bar at the bottom. Nothing runs on a schedule until you save.
Once you save the robot you'll be able to add things like monitoring, bulk runs, integrations, workflows, and export the data.
Adding columns to an existing table
You can add columns or whole new lists to a table you have already trained without creating a new robot.
Go to the Table you want to edit.
Click Add column.
Select data type, and click Open Design Mode.
Choose Single text, List, or Screenshot, then click Open Design Mode. Add Instructions if needed.
Click Preview, then Save Robot.
đ« If the robot uses clicks, inputs, or a password login, Table Studio canât add columns yet. Youâll be asked to retrain it in Robot Studio instead.
Reusing the data structure from another table
If you want to extract the same data from a different website that has a different layout/design, you can copy a data structure from another table.
From your Browse AI dashboard, click Build New Robot.
Enter the Origin URL: the page that shows the data you want.
Select Copy columns (schema) from existing table
Select the table you want to copy.
Click Start Training Robot.
Cleaning and enriching the data after extraction
Once the robot is extracting, you can add formula columns that post-process the data the robot captures: clean values, categorise records, or calculate new fields from the extracted data. Describe what you want in plain language and AI writes the formula for you.
Frequently asked questions
Will my existing robots break or change?
No. Table Studio changes how new robots are built. Existing robots keep running exactly as they did. You can retrain an existing robot in Table Studio if you want to.
Is an AI guessing at my data?
It proposes the tables and columns and shows you real extracted sample data; you review, refine, and approve before the robot is saved. Nothing runs on a schedule until you have saved it.
How does Table Studio handle pagination and long lists?
When the table includes a list, you choose how many items to extract at most. If the count looks like 12+, the preview was capped. Once you save, the robot extracts up to the list limit you set. To extract data across multiple pages, connect robots with workflows.
A column came back with the wrong data. What should I do?
Before saving, edit the columnâs name or Instructions to be more specific, then click Preview. Still stuck? Share your table URL with [email protected].
Does using Table Studio cost extra credits?
Once a robot is saved and running, it consumes credits like any other Browse AI robot.
Can I switch a Table Studio robot to Robot Studio later?
Yes. You can retrain any robot in Robot Studio at any time. The data in the robot's table is preserved.
Can I edit a Robot Studio robot in Table Studio?
You can retrain an existing robot in Table Studio if it doesnât use clicks, inputs, or a password login.