Build your first AI Browser Automation Workflow

You can adapt this workflow to almost anything. Swap the job title, location, websites, or output format to suit your own use case.

Build your first AI Browser Automation Workflow

You’re about to build an AI browser automation workflow using Opera Neon, OpenClaw, and Opera Browser CLI.

In around 20 minutes, you'll have an autonomous job search running inside Opera Neon. A browser-native AI agent handles everything from the search to the WhatsApp summary.

Our example scans three of Norway’s most popular job platforms—LinkedIn, Jobbnorge, and Finn—for newly posted product manager roles. It extracts the relevant details, saves them to a CSV file in Google Drive, and sends a summary via WhatsApp every day at 10 a.m.

Below you’ll see exactly how OpenClaw orchestrates the workflow, how Opera Browser CLI connects the AI agent to the browser, and how Opera Neon executes every step inside the browser session you're already using.

OpenClaw, Opera Browser CLI, and Opera Neon

This workflow has three moving parts.

• OpenClaw interprets your prompt and plans the workflow.
• Opera Browser CLI gives the AI agent a reliable way to interact with your browser.
• Opera Neon executes the workflow using the browser session you're already signed into.

After you connect everything, you just have to describe what you want in natural language.

The agent takes care of the rest.

Step 1: Add Opera Browser CLI to OpenClaw

The first thing we need to do is make Opera Browser CLI available as a skill inside OpenClaw.

Here’s how:

• Head over to the Opera Browser CLI GitHub repository and download the skill.md file.
• Next, open your OpenClaw workspace and navigate to the skills directory.
• Create a new folder called "Opera Browser CLI" and move the downloaded skill.md file into it. Refresh your skills list.
• You should now see Opera Browser CLI appear as an available skill. Activate it.

OpenClaw can now communicate with Opera Neon.

Step 2: Run your first AI Browser Automation prompt

With Opera Browser CLI installed, it's time to give your AI agent something to do.

Instead of writing browser automation code, you're simply going to describe the workflow in natural language.

Here's the prompt we're using:

### Prompt

Just one time Task: Automated Job Search and Data Extraction. Environment: Launch via Opera Neon using OpenClaw and Opera Browser CLI in a new window. Ensure sessions for LinkedIn, jobbnorge, and Finn remain authenticated.

Workflow: Search & Filter: Navigate to LinkedIn.com, jobbnorge.no, and Finn.no. Execute a search for “Product Manager” roles located in “Oslo, Norway”. Apply a filter for jobs posted “Within the past 24 hours”.

Data Extraction: Parse the top 5 results from each platform. Extract the following details for each listing:

Company Name Key Skills Required Salary Range, if explicitly visible Link Reporting & Integration: Append the extracted data to a CSV file hosted on Google Drive.

Send the direct link to the updated document to my WhatsApp.

Here’s what this workflow does:

• Open LinkedIn, JobbNorge, and Finn in Opera Neon.
• Look for product manager opportunities in Oslo, Norway.
• Filter listings posted within the last 24 hours.
• Extract the company name, required skills, salary (where available), and a direct link to each role.
• Append the results to a CSV file in Google Drive.
• Send a WhatsApp summary with the latest findings.
• Schedule the workflow to run every morning at 10:00 a.m.

Press Enter.

Now sit back and watch what happens.

What’s going on behind the scenes?

You've written one prompt, but behind the scenes it moves through a stack that turns your intent into individual browser actions.

Your prompt

↓

Large Language Model (understands your request)

↓

OpenClaw (plans the workflow)

↓

Opera Browser CLI (connects to the browser)

↓

Opera Neon (executes the actions)

↓

The web

The language model decides what needs to happen.

OpenClaw works out how to do it.

Opera Browser CLI gives the agent a reliable way to interact with Opera Neon.

Opera Neon carries out the workflow inside the browser session you're already signed into.

Now let's watch it happen.

Step 3: Watch Opera Neon execute the workflow

Hit enter and watch the plan turn into browser actions.

Opera Neon opens LinkedIn, Jobbnorge, and Finn, runs the searches, moves through the results, and pulls out the details you asked for. You don’t need to steer it through each page or keep feeding it instructions. Let it complete the run. Don't worry if it pauses for a few seconds now and then. Unlike a scripted automation, the agent is reasoning about what to do next before taking its next action.

This is where the difference becomes obvious… The agent isn't running inside a disposable automation browser. It's operating inside the Opera Neon session you already use, with access to your existing tabs, cookies, and signed-in accounts.

You can follow the whole thing as it happens: searches firing, pages opening, results being checked, and the useful bits being collected. The architecture from the previous sections isn’t a diagram anymore. It’s moving around in front of you.

In our run, the agent found eight matching roles.

Step 4: Review the results in Google Drive

Once the search is over, the browser work is done. The useful part is already waiting for you in Google Drive.

Open the CSV file, and you’ll see the roles collected with the company, key skills, salary (where provided), and a direct link back to the job post.

This is the bit that turns a clever demo into something you might actually use. Instead of opening three job sites manually, repeating the same search every day, and keeping 15 tabs alive while you compare listings, you’ve got one clean file you can scan in a few minutes.

In our run, Opera Neon found eight matching product manager roles and added them to the file automatically.

Step 5: Send the job search summary to WhatsApp

You don't have to watch the workflow finish. The WhatsApp message lets you know when there's something actually worth checking.

At the end of the run, OpenClaw sends the matching roles and a link to the results. Keep the message lean because you don’t want too much of the job data squeezed into a chat bubble.

You need just enough information to decide if the latest run is worth opening or not.

In our example, the message arrived with the eight new product manager roles and a direct link to the file.

Step 6: Schedule the AI Browser Automation workflow

Once the workflow has completed a clean test run, you’ll want to put it on a schedule. Ask OpenClaw to run the same job search every day at 10 a.m. Be explicit about the time zone, especially if the browser or machine might be running somewhere else.

Prompt: Run this workflow every day at 10:00 a.m.

Test first, schedule second.

An agent with a bad prompt and a daily trigger is just a punctual pain.

From here, the workflow runs without another prompt from you. Every morning, it checks for newly posted roles, updates these results, and lets you know when the run is complete.

You’re all set

Your workflow is now running end to end: searching the three job sites, collecting the useful details, saving them to Google Drive, and sending you a message when the run is finished.

If you haven’t already, customize it. Make the prompt yours.

Change the role, location, job sites, information you want extracted, and the time it runs.

You should also keep an eye on the first few results and tighten your search to remove anything that comes back too broad or not quite useful enough. That’s usually how a good automation takes shape. The first run proves the workflow works. The next few make it work the way you do.

Once it’s running, leave it alone. Just chose the next browser task you never want to do manually again.