Set Up an AI Agent to Automate Browser Actions in 7 Minutes
Launch Hermes/SwitchYard so the agent can open pages, fill out forms, and collect data on its own following your instructions
- 1.Go to the official SwitchYard website and install the Chrome extension. Click «Add to Chrome» and confirm the installation. After installing, open the extension by clicking its icon in the browser toolbar.
If the service is unavailable in your region, use a VPN to download the extension from the Chrome Web Store.
SwitchYard ↗ - 2.Sign up using a Google account or email. After logging in, enter a base instruction in the «System Instructions» field that defines the agent's role and constraints. Copy the ready-made prompt below and paste it into the settings field.
Do not enter passwords from personal accounts into the agent instructions - use a separate browser profile with no saved sessions.
SwitchYard ↗PromptYou are a browser automation agent. Your task is to navigate web pages, extract specific information, and fill out forms based on user instructions. Rules: 1) Always wait for page to fully load before acting. 2) If a page requires login, stop and ask the user. 3) Extract only the data specified in the task. 4) Never submit forms with payment information without explicit user confirmation. 5) Report each action taken in a summary at the end.
What this prompt doesThis prompt sets the agent's role: automate browser actions - open pages, collect data, fill out forms. You don't need to replace anything; this is the base setup. You can refine the rules for your own tasks. - 3.Create a new task (New Task) by clicking the corresponding button in the SwitchYard panel. In the «Task Description» field, describe the specific action the agent should perform in the browser. Use the template below, replacing the text in square brackets with your data.
Start with a simple task on a public website to test how the agent works before giving it access to work resources.
SwitchYard ↗PromptGo to [https://example.com]. Find all items matching the keyword [your keyword]. For each matching item, extract: title, price, and URL. Save the results as a table and display them in the summary.
What this prompt doesThe prompt is in English because the agent interface runs in English. Replace [https://example.com] with the website address, and [your keyword] with the word or phrase to search for. The agent will open the site, find the relevant items, and collect them into a table. - 4.Click «Run» to start the task. The agent will open a new tab, navigate to the specified website, and begin following the instructions. Watch its actions in the browser window - you will see each action in real time. If the agent stops with a question, answer in the SwitchYard chat.
Do not close the tab the agent opened for work - this will interrupt task execution.
SwitchYard ↗ - 5.When the agent finishes the task, check the results in the SwitchYard panel - a final report will appear there. If you need to repeat a similar task, clone the existing one using the «Duplicate» button and edit the description. For recurring tasks, set up a schedule in the «Schedule» section by specifying the frequency.
The free plan has a daily task limit - check current limits on the website before setting up recurring runs.
SwitchYard ↗
💰 What it costs to followprices as of 2026-08
- freeSwitchYard Free TierThe free plan gives a limited number of tasks per day - enough for testing and small recurring operations; exact limits are listed on the website.
- subscriptionSwitchYard ProThe paid plan removes task limits and adds scheduling; check the pricing page for current prices.
Try it and be the first to check in
Why today
AI Agent tutorialsyoutube
Recipe of the week by email
AI services change fast - interfaces and free limits may differ from what's described.