Generate ad concepts from existing research

High cost

Skip the research step — pass research text directly. Useful when you've already called research_url or have your own context.

How it works

Most flexible concept-generation entry point. Pass research + optional selectedIdeas (from generate_idea_seeds). Auto-saves to history.

Inputs

Name

Type

Required

Description

research

string

required

Pre-existing research text.

url

string

optional

URL to attach to the saved generation.

profileId

string

optional

Profile to apply.

brandId

string

optional

Brand context.

count

integer (1–10)

optional

Concept count. Default 5.

selectedIdeas

Array<{ title, description, note? }>

optional

Direct the concepts around specific ideas (often from generate_idea_seeds).

customInstruction

string

optional

Steer generation.

Returns

{ concepts: Concept[], generationId, mode }

Examples

Two-phase flow: ideas → concepts

Agent ran generate_idea_seeds, user picked 5 ideas. Now generate full concepts around them.

{
  "research": "<blob>",
  "count": 5,
  "selectedIdeas": [
    {
      "title": "Driveway Departure",
      "description": "..."
    }
  ],
  "brandId": "ectL5JsJKz55ItiFzEkB"
}

Related tools

generate_idea_seeds

Generate 10 creative direction seeds

Just the 'creative directions' phase — returns 10 short {title, description} pairs to feed into generate_concepts_from_research.

research_url

Research a URL with Perplexity

Run Perplexity sonar research on any URL — extracts business name, location, services, USPs, audience, hooks.

All tools Get API key