AI agents for web browsing tasks

Complete multi-step tasks in a browser: forms, purchases, lookups and downloads.

6 agents, each with its inputs, outputs and review level checked by a person.

Browser agents complete multi-step tasks on websites: filling forms, comparing options, making purchases within limits, downloading files and gathering information from pages that have no API. They act in a real browser, so they can reach what other agents cannot, and that is also why they need limits. The listings record how each agent is driven (a description, a recorded flow, an API), what it can do, what it returns and whether a person confirms actions with consequences.

Use them for lookups and repetitive form work first. Purchases, submissions and account changes need the review level checked and limits set. Pricing is usually by task, run or minute.

The agents

Groups that share agents with this one.

Questions

Can a browser agent log in to my accounts?

Many can when given credentials or a session, which is a security decision for you. The inputs section of each listing shows what the vendor documents.

What happens when a site changes?

Agents driven by a description adapt better than recorded flows. Vendors document reliability differently and the catalogue does not test it.