Web Search Research
Search and scrape current public web sources with Firecrawl, then synthesize evidence with nearby links.
Works with
Web Search Research
Search current public web content with Integration · Firecrawl and synthesize only claims supported by returned sources.
Search plan
- Restate the question, freshness requirement, country, language, and exclusions.
- Start with one precise query. Use additional queries only to close a known evidence gap.
- Inspect the live schema and select
limit,lang,country,formats, andtimeoutdeliberately. - Prefer Markdown output when the user needs content synthesis. Request other formats only when they are necessary.
Evidence handling
- Compare publication date with the date of the underlying event.
- Prefer primary sources for factual, technical, legal, financial, or medical claims.
- Keep source URLs next to the supported statement.
- Separate direct evidence, source claims, and your own inference.
- Report disagreement or missing evidence instead of forcing consensus.
Do not treat the provider's ranking or generated summary as independent verification. Do not search private systems, bypass access controls, or present stale content as current.
Retry policy
If the Action fails, report the returned error. Narrow an overbroad query or lower the requested result count only when that addresses the failure. Ask before a materially different paid search. Do not retry when the result says it is non-retryable.
Attribution
This native Loadout re-authoring retains the source package's search, deeper research, and content-extraction intent. Tavily credentials and shell execution were removed; the provider is now Firecrawl and results can differ. See and .