Brave Search research brief
Read-only Brave Search workflow for combining web results and LLM-ready grounding context for an explicit research query. It does not call Brave Answers, rich callback flows, POST variants, Goggles definitions, local POI detail endpoints, autosuggest, spellcheck-only routes, or arbitrary Brave API paths.
How it works
Inspect the data fetches, transforms, gates, and output this recipe runs.
Process (12 steps)
web
?
llm_context
?
web_results
default
Apply default
news_results
default
Apply default
video_results
default
Apply default
web_rows
slice
Take a subset of
web_count
count
Count items in web results
news_count
count
Count items in news results
video_count
count
Count items in video results
summarize
?
web_table
to_table
Format results as a data table
card
to_summary
Format results as a summary card
Settings
Configurable at install. Defaults shown — change them anytime in Recued.
count
setting
=
10
focus
setting
=
Summarize the strongest web results and LLM context for a research user. Call out useful sources, stale or thin evidence, and follow-up searches. Do not invent facts outside the Brave Search responses.
query
setting
=
country
setting
=
US
freshness
setting
=
row limit
setting
=
10
max length
setting
=
900
search lang
setting
=
en
brave search
setting
=
context urls
setting
=
8
context tokens
setting
=
8192
Trust & control
What installing this recipe would let it do. Recued grants these permissions at install — review them there before approving.