How is context processed and saved

This guide explains how Context Awareness works behind the scenesโ€”how page content is collected, stored, and refreshed to power intelligent agent behavior.

How context is processed

  • On the first load of the page where your agent is deployed, CustomGPT crawls the page content and generates a summary.
  • The first 500 words of the page are used to create that summary.
  • This summary is then saved and used in the agentโ€™s internal system prompt to help it understand user context.
๐Ÿšง

Note:

This is a premium feature.


How context is saved

  • Each agent can store content for up to 1,000 pages.
  • If the page is behind a login or requires authentication, we canโ€™t crawl it and context cannot be generated.
  • If no page-specific content is available, the agent will use the home page context as fallback.
  • If no home page content is available either, the agent will not be context aware for that session.
  • Saved content is automatically refreshed every 30 days to ensure the agent stays current with page updates.
โ—๏ธ

Note:

This feature is continually evolving and will become more flexible over time. If you have questions or want to request custom handling, please contact our support team.


Related articles