What problem does it solve? Automating real browser interactions—logging in, clicking elements, filling forms, and reading rendered pages—requires a controlled browser with persistent session state, which plain HTTP requests or web search cannot provide. ## Core Features & Use Cases - Managed Runtime Bootstrap: Verifies and installs the pinned agent-browser CLI (0.34.0+) and Chrome for Testing through a Vetta-managed npm prefix without touching the user's system installation. - Session-Scoped Automation: Runs every browser command under the task's VETTA_AGENT_SESSION_ID with --pin-tab, so concurrent agent tasks never share or hijack browser state. - Interactive Page Control: Uses snapshot refs (@e1) with click, fill, type, select, wait, and batch commands to drive dynamic pages, with account isolation via session key suffixes and --restore for persistent logins. - Use Case: Ask the agent to log into a dashboard, navigate to a report page, fill a filter form, and extract the resulting table text—all within an isolated, reviewable browser session. ## Quick Start Ask the agent to open a website in a real browser, fill in the login form, and read the resulting page content.
