Skip to main content

Browser and Sitemap Memory

Explore an Unfamiliar Surface

The agent starts from the current website. It inspects only the pages and UI needed for the requested outcome, reusing one Session for related work.

Learn While Browsing

As the agent discovers useful navigation paths and page structure, Webcmd can retain sanitized sitemap memory locally. Future tasks use that memory as a head start: where relevant content usually lives, how sections relate, and which routes were previously useful. This happens during normal browser work. Learning is invisible in normal output. Diagnostics appear on request, verbose mode, or a material warning-retention failure. You do not need to ask the agent to build or maintain anything separately.

Adapt When a Site Changes

The live page is always the source of truth. Before acting, the agent checks that remembered structure still matches the current website. If a route, control, or layout has changed, it explores the new surface and refreshes the useful knowledge rather than forcing an outdated path.

What Memory Does Not Store

Sitemap memory is for reusable site structure, not secrets. Credentials, one-time codes, cookies, and private page contents should not be placed in prompts or retained as navigation knowledge. See Privacy for storage, provenance, and seed-lookup details.

Ask for the Right Outcome