How to Do Microsoft Scout Web Scraping With Playwright

Microsoft Scout web scraping lets you automate browser tasks like extracting tables, filling forms, and capturing screenshots just by describing what you need, without writing a single line of Playwright code. This guide walks through enabling access, setting permissions, and running your first scraping task with Scout.

What Is Microsoft Scout and Headless Web Scraping

Scout is an always-on AI agent Microsoft introduced for Microsoft 365 at Build 2026. It drives a real Playwright-controlled browser on your behalf, so you type a task in natural language and Scout handles the clicking, scrolling, and data extraction.

This is a form of headless web scraping, where a browser runs without a visible interface. Playwright controls Chromium, Firefox, and WebKit in this headless mode, which lets automation run faster and at scale compared to a browser window you have to watch.

Scout reaches Playwright through the Model Context Protocol (MCP), the same interface that lets AI models such as Copilot, GPT-4, and Claude control browser automation without custom integration code. Scout runs in two modes depending on the task: snapshot mode reads the page through the accessibility tree, and screenshot mode is used when the AI needs to understand the visual layout instead of just the structure.

What You Need Before You Start

Scout is a Frontier preview feature, so it is gated in a few layers before it will actually run. Before touching any settings, confirm you have the following, since missing any one of these is the most common reason sign-in fails later:

A Microsoft 365 Copilot license on your account.
A GitHub account, since Scout uses it for sign-in and token billing.
An admin who can act on your tenant, since two of the steps below cannot be done from a standard user account.

With those in place, setup happens in two stages: an admin enables Frontier and Scout at the tenant level, then you install the app and turn on permissions yourself.

Method 1: Enable Browser Control for Scout

Browser control is not turned on by default. This is the admin-side setup, so if you are not an admin on your Microsoft 365 tenant, send these steps to whoever is.

Step 1: Enroll the Tenant in Frontier

Sign in to the Microsoft 365 admin center, select Copilot in the left navigation, then Settings, then View all. Search for Frontier and select Copilot Frontier. Set access for the organization, either all users or specific groups, and select Save. This change can take up to about three hours to take effect, so do not expect Scout to appear immediately.

Step 2: Complete the Scout Attestation Form

Because Scout can send prompts and data to GitHub Copilot for processing, Microsoft requires a separate opt-in for this beyond the Frontier enrollment above. An admin needs to open the M365 Admin – Microsoft Scout Sign-up Form and accept the terms on behalf of the organization. Skipping this step is a common reason Scout stays inaccessible even after Frontier is turned on.

Step 3: Turn On Browser Control in Agent Management

This is a separate toggle from the Frontier enrollment in Step 1, and both are needed. Go to the Microsoft Admin Center, open Agent Management, and enable Scout’s browser control feature for your organization.

Step 4: Install Scout and Sign In

Once the admin steps are done, the end user downloads and installs the Microsoft Scout desktop app, then signs in with work credentials. Sign-in only succeeds once both admin gates above are complete, so if sign-in fails, check with your admin before assuming the app is broken.

Step 5: Enable Permissions Inside Scout

Open Scout, go to Settings > Permissions, and toggle on Browser Control and Web Browsing. Give it a minute before running a task, since permission changes do not always apply instantly.

Method 2: Run a Scraping Task With Scout

Once browser control is enabled, you do not need to write any code. You describe the task directly in the chat.

Step 1: Open a Scout Chat and Describe the Task

Open the Scout desktop app and start a new chat. Type your request in plain language. Examples include:

Go to the Azure portal and check my resource group’s cost this month.
Take a screenshot of the pricing page.
Extract the table on this page and turn it into a CSV.
Fill out the contact form with this information.

Step 2: Let Scout Drive the Browser

Scout opens a Playwright-controlled browser session, navigates to the page, and performs the steps needed to complete your request, whether that is reading a table, clicking through a form, or capturing a screenshot. Scout shows its progress in the chat as it works, and for actions it flags as sensitive, such as submitting a form, it can pause and ask for your approval before continuing rather than completing everything unattended.

Step 3: Review the Output

Check the extracted data, screenshot, or completed form before using it further. Automated extraction can misread dynamic layouts or miss fields that changed since the last page update.

Limitations to Keep in Mind

Scout is still a preview tool, so a few restrictions are worth knowing before you rely on it for regular tasks.

Scout is still a Frontier preview feature, so access depends on enrollment and administrator settings.
Administrators can block Playwright, restrict permissions, or require approval before Scout performs certain actions.
Login walls, CAPTCHAs, and frequently changing page layouts can interrupt or break automated runs.

Frequently Asked Questions

Do I need to know how to code to use Scout for web scraping?

No. Scout is built around natural language prompts, so you describe the task and it drives the Playwright-controlled browser without you writing any automation scripts.

Why can’t I access Scout’s browser control feature?

Browser control depends on being enrolled in the Frontier Preview Program and having it enabled through the Microsoft Admin Center. If either step is missing, or your administrator has restricted it, the feature will not be available.

Can administrators block Scout from visiting certain websites?

Yes. Administrators can block specific browser destinations, disable Playwright entirely, or require manual approval before Scout carries out an action.

Does Scout work with CAPTCHAs and login pages?

Not reliably. Login challenges and CAPTCHAs can interrupt automated browser sessions, so tasks involving authentication may need manual steps.

Why do I need a GitHub account to use Scout?

Scout uses GitHub Copilot to process prompts and responses, and your GitHub account is used for sign-in and token billing. Without one linked, sign-in will not go through even after your admin completes the tenant-level setup.

Related Scout Setup and Admin

Scout Features

Scout for Developers