Introduction

The in-person step your sales motion is missing

Your AI agent can research a thousand leads before breakfast. It can write personalized outreach, qualify prospects, and schedule calls. But the moment a deal requires trust — a handshake, eye contact, a shared coffee — the agent hits a wall.

That wall is the most valuable part of the sales funnel. The customers who close through real human connection stay longer, spend more, and refer others. They have the highest lifetime value. And no email sequence can replicate that.

HumanConnection is the in-person layer for AI-powered sales. It gives your agent the tools to find the right person, understand who they are, discover their interests, find where they’ll be, and orchestrate a face-to-face meeting that turns a lead into a long-term relationship.

Why in-person matters

The data is clear: in-person interactions create disproportionate business value.

Cold emailVideo callIn-person meeting
Trust formationNoneLowHigh
Close rate1-3%10-15%40-60%
Average deal sizeSmallMedium2-3x larger
Customer lifetime valueBaseline1.5x3-5x
Referral likelihoodRareOccasionalCommon

The reason is simple: humans are wired for face-to-face connection. A 20-minute coffee creates more trust than 20 emails. A shared event creates a sense of belonging that no Zoom call can replicate. And that trust compounds — into renewals, upsells, and referrals that multiply LTV.

HumanConnection exists to make this the default, not the exception.

What is HumanConnection?

HumanConnection is an open-source toolkit that bridges AI agents and the real world. It provides tools organized around four capabilities:

1. Find people

The core tool, search_humans, supports two modes. Person lookup takes a name and optional company or location info and returns the person’s LinkedIn, Instagram, X/Twitter, and TikTok profiles, their bio, location, email, and a confidence score. People search takes a location and title filters (with AND/OR/NOT boolean syntax) to find professionals matching specific criteria. Under the hood, it runs parallel searches across the open web using SearXNG, crawls Linktree pages for handle discovery, and scores candidates using keyword similarity.

2. Discover interests

search_interests takes a person’s name and social profile URLs, then searches their recent activity on LinkedIn, X/Twitter, Instagram, and the web for personal interests, hobbies, and conversation starters. These signals help your agent pick the right event or venue and personalize the meeting.

3. Find meeting opportunities

search_events discovers conferences, meetups, and trade shows near a lead using the Luma discover API. search_places finds restaurants, cafes, golf courses, tennis/padel courts, galleries, museums, stadiums, and bars. Events are the highest-conversion meeting context; places are the always-available alternative.

4. Orchestrate the meeting

send_meeting_invite sends a Google Calendar invite to all participants via email — the invite is the booking. After the meeting, send_meeting_report captures the transcript, outcome, and next steps, and send_meeting_payment rewards the human who made it happen.

How it works

1

Search for the lead

Your agent calls search_humans with a name and company domain (e.g. extracted from the lead’s email). HumanConnection searches across LinkedIn, Instagram, X/Twitter, TikTok, and the open web. The best match per platform is returned with a confidence score, location, and optionally their email.

2

Discover their interests

Use search_interests with the lead’s name and discovered social URLs. HumanConnection scans their recent LinkedIn posts, tweets, and Instagram activity for personal signals — hobbies, sports, causes, travel. These signals drive event and venue selection.

3

Find where to meet

Map interests to event categories and place types. Use search_events to find conferences, meetups, and trade shows near the lead. Use search_places to find restaurants, cafes, or other venues. Event-based meetings feel natural — you’re not asking for their time, you’re both already there.

4

Find a sales rep nearby

Use search_humans in people search mode with the lead’s location and title filters (e.g. "sales AND SaaS") to find available representatives in the area. Use provide_email to get their contact info.

5

Book and invite

Use book_ticket to register for the event if needed, then use send_meeting_invite to send a Google Calendar invite to both participants. Everyone gets the time, place, and agenda in their calendar.

6

Get the report

After the meeting, send_meeting_report captures the summary, outcome (deal_closed, follow_up_needed, not_interested, no_show), deal value, and recommended next steps.

7

Reward the human

Use send_meeting_payment to pay the human based on the outcome. Commission, flat fee, or bonus — aligned incentives mean better conversations.

Real results

Here’s what the full pipeline produces across 5 different industries — from lead discovery to interest signals to matching events, venues, and sales reps:

E2E sales motion results across 5 industries

Quick example

1import { search_humans, search_interests } from '@humanconnection/toolkit';
2
3// Find a person using their company domain
4const results = await search_humans({
5 name: "Brian Chesky",
6 company_domain: "airbnb.com"
7});
8
9console.log(results);
10// [
11// {
12// "name": "Brian Chesky",
13// "email": "brian@airbnb.com",
14// "company_domain": "airbnb.com",
15// "bio": "Co-Founder & CEO at Airbnb.",
16// "location": "San Francisco, CA",
17// "profiles": {
18// "linkedin_profile_url": "https://linkedin.com/in/brianchesky",
19// "x_profile_url": "https://x.com/bchesky",
20// "instagram_profile_url": "https://instagram.com/bchesky"
21// },
22// "confidence": 0.85
23// }
24// ]
25
26// Discover their interests for meeting personalization
27const interests = await search_interests({
28 name: "Brian Chesky",
29 linkedin_url: "https://linkedin.com/in/brianchesky",
30 x_url: "https://x.com/bchesky"
31});
32
33console.log(interests);
34// {
35// "signals": [
36// { "source": "linkedin", "signal": "Posted about design thinking and product craft" },
37// { "source": "x", "signal": "Frequently discusses travel industry trends" },
38// { "source": "web", "signal": "Known art collector and design enthusiast" }
39// ]
40// }

Or use the CLI:

$npx @humanconnection/cli humans search --name "Brian Chesky" --company-domain "airbnb.com"

Available tools

ToolDescriptionStatus
search_humansSearch for a person by name (with company/location context) or find people by location and title filters. Returns social profiles, bio, location, email, and confidenceLive
search_interestsSearch a person’s recent social activity for personal interests, hobbies, and conversation startersLive
search_eventsSearch for conferences, meetups, and trade shows near a locationLive
search_placesSearch for restaurants, cafes, golf, tennis/padel, galleries, museums, stadiums, barsLive
search_ticketsFind where to buy tickets for events across ticketing platformsLive
book_ticketRegister for an event on behalf of someone (Luma, Meetup, Eventbrite)Live
send_meeting_inviteSend a Google Calendar invite to all meeting participantsPlanned
send_meeting_reportSubmit a post-meeting report with transcript, outcome, deal value, and next stepsPlanned
send_meeting_paymentPay a human for a completed meeting (commission, flat fee, or bonus)Planned

Packages

PackageDescription
@humanconnection/toolkitCore library with tool definitions, Zod schemas, and handler functions
@humanconnection/cliCommand-line interface

Hosted services

EndpointDescription
api.humanconnection.sh/api/v0/*REST API — POST to any tool endpoint
mcp.humanconnection.sh/mcpMCP server via Streamable HTTP

Infrastructure

HumanConnection runs two key services:

  • SearXNG — A self-hosted meta-search engine that queries Google, Bing, DuckDuckGo, and other engines through a DataImpulse residential proxy for reliable results. Each query uses a single engine (Google or Brave) to reduce load and improve reliability.
  • Crawl4AI — A web scraper for Linktree crawling, profile verification, and bio extraction.

Next steps