BROWSER INFRASTRUCTURE · FOR AI AGENTS
Oxylabs Agent Browser
Control remote Chromium sessions with managed proxy access.
Use CDP-compatible automation or an MCP integration to navigate dynamic pages, collect content and inspect remote browser sessions.
Your integration checklist
- Capabilities
- What you need from your owner
- Create an Oxylabs account, enable Agent Browser credentials and set permitted targets, session concurrency and a traffic budget.
- Authentication
- Connect to hb.oxylabs.io over secure WebSocket using the full instance username, including its suffix, and password in the connection URL. Keep that URL out of logs.
- Your first task
- Open one permitted dynamic page, extract a known field, save a screenshot and close the remote session.
- Billing & payment
- Subscription credits pay for browser traffic measured in GB. Plan rates differ, so large media downloads and unnecessary page resources can materially increase cost.
- Agent payment support
- Not verified. Programmatic usage does not establish autonomous checkout support.
- Operational limits
- The documented baseline is 100 concurrent sessions and 10 session starts per second. Site behavior and account access still affect results; a remote browser does not guarantee task success.
ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK
Prove the interaction before scaling browser sessions
Editorial recommendation · Reviewed 2026-10-16
When to choose this service
Choose this service for a workflow that needs interactive rendering and managed browser or network capacity. A static page that already returns usable HTML may need only a simple fetch.
The dependency is operated remote browsers and an external proxy network. Generating a Playwright script does not provide those running sessions or network resources.
Suggested workflow
- Define the expected page state, permitted interaction and output fields. Start one session with the required region and viewport.
- Run the smallest representative interaction and compare the extracted result with a screenshot or page state. Keep authentication material outside saved artifacts.
- Expand only after the sample succeeds. Cap concurrent sessions and downloaded resources, and close each session in a cleanup handler.
Acceptance checks before real work
- Confirm the page is the intended destination rather than an error or login screen.
- Verify the final page state before repeating any action that could submit data.
- Check that browser sessions close on both success and failure.
If the workflow fails
After a timeout, inspect the current page state before replaying a click or submission. Resume from the last verified step and reduce concurrency when account limits are reached.