ADVANCED · 8 MODULES · 64 LESSONS

Computer Use and Browser Agents

Build an agent that works in systems with no API the way a person would: it reads the screen, clicks, types, checks the result, and asks before anything irreversible.

For you if
You already built an agent with tools and need it to operate portals or legacy systems.
By the end
You'll know how to sandbox it in a container, avoid misplaced clicks, and defend it from on-screen text that gives it orders.
You need
Intermediate Python, basic HTML, and Docker. No API key needed.
  1. 01What computer use is and when to use it8 LESSONS
    1. Introduction: a portal with no API
    2. Three ways to drive an interface
    3. The observe-act-verify loop
    4. Provider tools in November 2026
    5. Lab setup, part 1: sites and containers
    6. Lab setup, part 2: driver, simulated pilot and loop
    7. The first run
    8. Project: decide the approach for four tasks
  2. 02Playwright for agents8 LESSONS
    1. Introduction: the script the agent has to beat
    2. Browser, context and page
    3. Locators versus coordinates
    4. Auto-waiting and slow pages
    5. Screenshots and what is on top
    6. Network routes and dialogs
    7. The baseline script
    8. Project: make the baseline survive a redesign
  3. 03Provider computer use tools8 LESSONS
    1. Introduction: the same drone, three radios
    2. The Claude computer toolset
    3. Batches, halt text and errors
    4. Writing a driver for the members
    5. The OpenAI computer tool
    6. Computer use in Gemini
    7. The SDK toolset classes and the tool runner
    8. Project: add a member to the driver
  4. 04Screens, coordinates and scaling8 LESSONS
    1. Introduction: the photo is not the terrain
    2. CSS pixels, device pixels and screenshots
    3. Image limits and visual tokens
    4. Resizing and scaling coordinates back
    5. Refusing points outside the screen
    6. Choosing the resolution
    7. Zoom and small text
    8. Project: a scaling test suite
  5. 05Observe, act and verify8 LESSONS
    1. Introduction: the pilot who assumes
    2. Why models assume outcomes
    3. Detecting change on screen
    4. Checking the expected state
    5. Retries, limits and fallbacks
    6. Verify the outcome in the system of record
    7. Screenshots in the context window
    8. Project: an in-loop check
  6. 06Browser agents with element refs8 LESSONS
    1. Introduction: the map with labels
    2. The Claude browser toolset
    3. read_page and the page tree
    4. Acting by ref: form_input and clicks
    5. Stale refs
    6. Refs or pixels: when to use each
    7. Tabs, state and navigation
    8. Project: the same task in both modes
  7. 07Sandbox: Docker and a virtual desktop8 LESSONS
    1. Introduction: the fenced flight zone
    2. Why the agent runs in a container
    3. Hardening the browser container
    4. A virtual desktop with Xvfb
    5. The desktop driver
    6. Egress and the loopback trap
    7. Files, credentials and the browser profile
    8. Project: run the desktop task in the sandbox
  8. 08Screen injection and human confirmation8 LESSONS
    1. Introduction: a sign painted on the ground
    2. Prompt injection from the screen
    3. Layered defenses in practice
    4. What counts as consequential
    5. Designing the confirmation
    6. Approvals on a desktop
    7. The audit trail
    8. Final project: the full run with injection and approvals
Reviews

What students say

These reviews are from enrolled students who completed at least 50% of the course. We moderate reviews only on content grounds (spam, offensive language, personal data), never for being critical or negative.

No approved reviews yet.

Be the first to share your experience!

SEC. 06 · THE STUDIO

Why this studio exists

For decades, building software belonged to technical profiles with years of training, and getting a digital product into the real world took companies months or years. No-code and AI have narrowed that gap, but the knowledge is still held by a few.

Tech Benders exists to open that knowledge up, so business teams and non-technical teams can put their ideas into production with the right technical decisions behind them. AI does what you ask of it. Without a technical base, what it builds may not scale with your company’s ecosystem, and its costs can run away over time.

The team knows the pains of large organizations and the speed at which they have to move. It brings solution architecture, data engineering, product built end to end from back end to front end, and daily work with AI. Software that solves what you need in days, not months, with the cost of AI kept in check.

Founders

The studioRead about Tech Benders →

Write to ushi@techbenders.io

Made with Wislab