Computer Use and Browser Bridge
Axiomate can do more than read code and run commands. When needed, it can connect to local desktop and browser capabilities. The key idea is not just “look at an image and click”. Axiomate can combine local element detection, window information, control positions, and input events so even AI models without image-recognition support can still perform some computer-control tasks.
For quick start, understand two concepts first:
- Computer Use: lets Axiomate operate the local desktop environment, including screenshots, mouse, keyboard, windows, clipboard, and local elements.
- Browser Bridge: lets Axiomate connect to an isolated local Chromium browser through the browser debugging protocol, inspect pages, click elements, fill forms, read console output, and take screenshots.
Both let Axiomate see and operate external interfaces, but they fit different use cases. Computer Use is especially useful because it uses local capabilities to compensate for model limits: even if the current AI model cannot recognize images directly, Axiomate can still use local elements and operation tools to understand interface structure and perform actions.
When to use Computer Use
Computer Use is better for desktop-level operations, such as:
- Inspecting the current screen or an application window.
- Operating a desktop app that has no command-line interface.
- Using keyboard, mouse, or clipboard in a workflow.
- Checking installers, system dialogs, or native GUI state.
Computer Use works against the real desktop environment, so its actions are close to what a user would do manually. Be extra careful around login, sensitive information, external submissions, or irreversible actions.
When to use Browser Bridge
Browser Bridge is better for web and frontend debugging, such as:
- Opening a local preview site and checking page content or interactions.
- Reading the page accessibility tree and locating elements by button, text, label, or role.
- Reading browser console output to find frontend errors.
- Taking screenshots to inspect layout, popovers, images, and responsive behavior.
- Filling forms or clicking links to validate user flows.
Browser Bridge requires a Chrome/Chromium-based browser installed locally. Installing Google Chrome is recommended.
Browser Bridge uses an isolated Chromium profile. It does not take over the user's real browser session or extensions. This makes it a good default for safely debugging local pages and public websites.
How to choose
| Scenario | Prefer |
|---|---|
| Check website pages, frontend interactions, or console errors | Browser Bridge |
| Inspect a local app window, installer, or system dialog | Computer Use |
| Find web buttons, inputs, or links by semantic locators | Browser Bridge |
| Operate a non-web GUI | Computer Use |
| Only read code or run tests | Usually neither |
Relationship with permissions
Computer Use and Browser Bridge expose capabilities that are closer to real user actions. Tool calls still go through Axiomate's current permission mode and approval flow.
Recommended practice:
- Ask Axiomate to explain which window or page it is going to operate.
- Confirm actions involving login, payment, sending messages, submitting forms, or deleting data.
- For web debugging, prefer Browser Bridge over controlling the user's real browser.
- If the task only requires reading code or running tests, do not use desktop or browser automation.
Quick examples
To check a local preview page, you can say:
Open the local preview page, check whether the new quick-start article appears in the sidebar, and inspect the browser console for errors.
Axiomate will usually prefer Browser Bridge because this is a web debugging task.
To inspect an installer window, you can say:
Look at the current installer window and confirm whether the Next button is available.
This usually requires Computer Use because the target is a desktop window, not a web page.