Glossary term

Browser agent

An agent that operates a web browser — navigating, reading pages, filling forms and clicking — either as a product of its own or a coding agent with a browser tool.

Definition

A browser agent is an AI agent whose tools are a web browser: open a URL, read the page or its accessibility tree, type into fields, click, upload files. It can be a dedicated product or a coding agent extended with a browser integration. A headed browser is one you can see; a headless one runs invisibly.

For form work the important properties are which browser profile it uses (your signed-in accounts or a blank one), whether you can watch it, and whether a successful tool call is verified against the page.

Why it matters for a directory launch

Directory forms live behind logins, verification emails, CAPTCHAs and two-factor prompts. A browser agent in your own signed-in Chrome can do the repetitive part while you approve the connection, solve the challenge and keep control of the accounts. A headless agent with your credentials somewhere else is a different risk profile entirely.

How LaunchRepo handles it

BROWSER-SETUP.md connects the agent to the Chrome you already use — remote debugging enabled via the inspect page on Chrome 144+, or a dedicated local profile on a loopback port — through Chrome DevTools MCP. A smoke check on a public page precedes any submission. The rules: close unrelated sensitive tabs, no cookie export, no profile copying, a CAPTCHA or 2FA is a human handoff, and a tool call that returned is not proof the page saved.

See the term in the repository

LaunchRepo is 346 directory playbooks, a product context file and a submission tracker your own agent runs. Pay €99.98 once; use it for every product you own.

See pricing