Interests

Overview

search_interests scans a person’s recent social activity across LinkedIn, X/Twitter, Instagram, and the open web to find personal signals — hobbies, passions, causes, and topics they engage with. These signals help your agent pick the right event or venue for a meeting, and personalize outreach.

Without interests, you’re guessing. With them, you’re matching.

InterestSignal type

1
interface InterestSignal {
2
  source: "linkedin" | "x" | "instagram" | "web";
3
  signal: string;  // Raw snippet from the source
4
}
5
6
interface InterestsResult {
7
  signals: InterestSignal[];  // Up to 20 signals
8
}

Search parameters

Pass the social URLs discovered by search_humans — the more profiles you provide, the richer the signals.

Example

1
const interests = await search_interests({
2
  name: "Sara Blakely",
3
  linkedin_url: "https://linkedin.com/in/sarablakely",
4
  x_url: "https://x.com/SaraBlakely",
5
  instagram_url: "https://instagram.com/sarablakely"
6
});

1
{
2
  "signals": [
3
    {
4
      "source": "x",
5
      "signal": "Mom of 4. Inventor of @SPANX. Believe it's my calling to support women..."
6
    },
7
    {
8
      "source": "linkedin",
9
      "signal": "Sara Blakely's Post: #founderstory — How I started Spanx with just one product..."
10
    },
11
    {
12
      "source": "instagram",
13
      "signal": "Photo by Sara Blakely on April 02, 2026. May be an image of yoga."
14
    },
15
    {
16
      "source": "web",
17
      "signal": "Inducted into National Inventors Hall of Fame in 2026. Philanthropy: Red Backpack Foundation..."
18
    }
19
  ]
20
}

How it maps to events and places

The agent reads the signals and decides what to search for. This is reasoning, not keyword matching — the agent understands the person:

Real example: Sara Blakely

From 20 signals found, the agent maps:

• Entrepreneurship + women empowerment → search events with category startup → finds “Sisters Who Startup” in Seattle
• Yoga + wellness → search places with type cafe for a calm meeting spot
• Creative/design sensibility → search places with type gallery → finds A/NT Gallery, Vermillion

The meeting becomes natural — not a cold pitch at a random Starbucks, but a conversation at an art gallery after a founders event.

How search works under the hood

search_interests runs 4 parallel queries (one engine each to stay light):

1. X/Twitter: from:{handle} site:x.com → recent tweets and bio
2. LinkedIn: "{name}" site:linkedin.com/posts → recent posts and articles
3. Instagram: {handle} site:instagram.com → recent photos and captions
4. Web: "{name}" hobby OR passion OR yoga OR golf OR tennis... → mentions across the web

Results are deduped, filtered for boilerplate, and capped at 20 signals.

Relations to other resources

Best practices