Skip to main content

​
What it’s for

Scout is the public research agent. When you need to know something that lives outside your org — a company, a market, a product, a public figure, a recent article — Scout is who you talk to. It’s not the right agent for:
  • Touching your CRM (that’s Marcus or Athena)
  • Making decisions or acting on behalf of the company (that’s Athena)
  • Talking to customers (Marcus, or the leader’s Digital Twin)

​
How it works

1

You describe what you want to know

Examples:
  • “Research Magazine Luiza — business model + team size”
  • “Who are the top 3 competitors of Linx?”
  • “Has Stone’s CEO made any public statements about crypto in 2026?”
2

Scout searches the web

The first choice is the Tavily engine (grounded public-web search). If Tavily returns nothing useful — or if your org doesn’t have a key configured — Scout automatically falls back to a built-in web search as a secondary option.
3

Scout assembles the answer

The response comes with cited snippets: every claim is linked to the URL that was its source. No citation = Scout is not asserting that point (it appears as an explicit hypothesis).
4

You can ask for a deeper dive

“Go deeper on item 2” or “Read the full article 1” — Scout fires a read_url against the specific link and comes back with the unpacked content.

​
Scout’s tools

ToolPurpose
Web searchPublic web research (primary engine + automatic fallback)
URL readerReads the content of a specific URL in depth
Structured scrapingExtracts data from complex pages when search isn’t enough
Pain-point miningExtracts repeated patterns from public Google Maps reviews
Site evaluationChecks whether a URL matches an ICP / persona for your org
Every call generates a trace and an entry in the org’s usage history — auditable via “View trace” in chat or in the Billing panel. The cost of each tool is charged in Stars according to the current pricing table in the panel.

​
Rate limits

To protect your org’s key from accidental exhaustion:
  • Per-org cap — a searches-per-second ceiling is enforced per organization. When it’s hit, Scout automatically falls back to an alternative engine for a short period, then resumes.
  • Cap-fire audit — every time the cap fires, it’s recorded in the org’s event history so you can track it.

​
Examples of effective prompts

​
Company research

Research the company Petrobras:
1. Business model
2. Top 3 products by revenue
3. Recent ESG initiatives (past year)

Answer in short bullets. Cite the URL for each item.

​
Competitive analysis

Who are the 3 main competitors of Pipefy today?
For each, include: ICP served, average price, main
technical differentiator.

​
Fact verification

This article claims Stone had a 40% drop in Q1/2026.
URL: [link]
Confirm or refute with alternative sources.

​
Review mining (pain-points)

Mine pain-points from public Google Maps reviews of the
10 largest gyms in São Paulo. Focus on repeated patterns.

​
When Scout fails (and what to do)

SymptomLikely causeWhat to do
Vague response / no citationsSearch engine key not configuredSet the vendor key in the integrations panel
Recurring “I found nothing”Query too broad or wrong languageRephrase in specific pt-BR OR in English
Cap fires for minutesSearch loop in the chatPause + review the prompt — Scout doesn’t normally loop on its own
Cites example.com as a sourceTest (dry-run) mode enabled in the environmentCheck with the org admin — in production the default is already off

​
Next steps

Search engine

How to connect your public search engine key.

Marcus — outbound

When Scout finishes the research, Marcus takes over and writes the commercial outreach.

Athena — operator

For research + action in the same turn, Athena usually orchestrates better than Scout alone.

Costs

Each Scout call charges Stars from the org’s wallet.