Research information on websites

Lets the AI read web page content during support and use the text (as markdown) to answer with up-to-date information — limited to the sites you allow.

When to use:

• Look up catalog, inventory, pricing, or policies on the company site
• Follow docs, changelogs, or product pages without pasting text into the prompt
• JavaScript sites (Firecrawl engine) or static HTML (Plain HTML engine)

Configuration:

1. Add the Research information about… action
2. Choose the engine:
   • Firecrawl — requires the integration (organization API key); renders JS and allows optional free search
   • Plain HTML — static fetch, no key; best for non-SPA pages
3. Register the sites (URL + description of what is on each)
4. With Firecrawl: optionally enable free web search
5. Save the prompt

Behavior:

• The AI can change query params (e.g. ?search=Civic → ?search=Corolla)
• It can open subpages under the same path as the configured site
• If the page is only an index, the result may include followUpLinks for a second call

HTML engine

Sites built only with React/Next may return empty or incomplete results with Plain HTML. Prefer Firecrawl.

Security

With free search off, the server only accepts URLs from configured sites (and subpages under the same path).

Example:

Customer: Do you have a 2022 Corolla in stock?
AI: [scrapes the site with ?search=Corolla] Yes — I found 3 units…

Changelog: v2026.7.13 · Integration: Firecrawl