The problem your project solves

Every week I meet 100+ people across events, Telegram groups, Gmail inbounds, and business card exchanges — and easily double that at conventions. By Friday, I've forgotten most of them. And I'm not alone — every founder, BD person, partnerships lead, community builder, and private banker in SEA is drowning in the same problem.

Business cards pile up on desks in stacks we'll "deal with later" — until later becomes next month and all context is gone. Who was the one building infrastructure? Who wanted to speak at our next event? Who was from Tokyo and looking to expand? A name and title on a card tells you nothing you can act on.
Telegram and WhatsApp messages bury real relationship signals inside walls of noise. Someone sends Hi!! with genuine excitement and it gets missed among fifty other pings. Someone goes quiet for 60 days and the drift only registers when they finally stop replying altogether. The warmth of a relationship is invisible — scattered across a hundred threads, dozens of groups, and the emotional undertones of the way people punctuate.

Partnership emails need judgment — is this a real fit or polite spam from someone who found our LinkedIn? — and we rarely have the bandwidth in the moment to qualify, enrich, and reply in our own voice. So they sit. Days turn to weeks. By the time we respond, the counterparty has moved on to someone who answered faster.
Warm intros are the work that compounds relationships the most — and the work we do the least. We mean to connect two people who should meet. We forget. The moment passes. Both relationships weaken because we quietly failed to add value to either.

Post-event follow-ups have a 24-hour window before the momentum evaporates. Miss it and a warm conversation at TOKEN2049 becomes a cold LinkedIn connection three months later.

Cold relationships quietly pile up — hundreds of people we used to matter to, now sitting dormant in a graph nobody maintains. Every one of them was a live opportunity once. Most of them could be reactivated with a single honest message. Almost none of them will be.

The deeper problem: every CRM on the market — Salesforce, HubSpot, Attio, Dex, Folk — fails the same way. They ask you to stop working to log your work. So nobody logs. The graph is always wrong. Relationships always leak.

And at enterprise scale it's worse: when a senior BD person with 2,000 relationships in their head quits for a competitor, the company loses years of compounded trust in a single day. The "CRM will preserve institutional memory" promise has failed for three decades because employees don't log.

Connect Crew fixes this.

How you are solving it

Connect Crew is a department of six AI Crew that runs in the background across every channel you already use — Telegram, WhatsApp, Gmail, LinkedIn, business cards, voice notes, events. Not an agent. A department. Six specialists, one shared memory, one daily ritual.

The six Crew each own a specific problem:

Scan turns business cards into enriched contacts. Photograph one card or a whole stack fanned out on a table — Scan extracts every field, pulls each person's LinkedIn, their company's funding and size, and flags mutual connections across your existing graph. The stack goes from unprocessed guilt to a queryable relationship layer in under a minute.

Secret Signals reads how you already type — without anyone knowing. Hi!! registers as hot. Hi. as cooling. A two-minute reply, a weekend evening message, an unprompted voice note, a sudden 30-day silence — these are behavioural cues your brain already processes unconsciously. Secret Signals makes them legible by computing a live Warmth Index (0–100) per contact across 11 signal categories: punctuation, greeting style, emoji, response time, channel choice, timing, silence patterns, voice, and more. The other party never sees it. You finally do.
Inbound Connections qualifies cold partnership emails and drafts replies in your voice, citing your past deals and organisational context. Real fit or polite spam? The Crew decides, drafts, and stages. You just approve.

Warm catches post-event relationships before the 24-hour window closes. Met someone at TOKEN2049 last night? Warm has the follow-up drafted before morning coffee, referencing what you actually talked about.
Cold surfaces relationships 60+ days silent and drafts honest resurrection messages — no forced apologies, just a real door reopening.

Scribe turns voice notes and meetings into structured memory: extracted commitments, named people, action items. What you said you'd do, captured so it doesn't die in a Plaud recording you never replay.
All six share one Relationship Graph — your property, portable, exportable, deletable. Every signal, every interaction, every commitment feeds into one memory. Claude handles the reasoning and writing. Genspark Super Agent handles the web enrichment. The handoff between them is visible in the UI in real time, so you can always see the Crew working.

Every morning at Morning Connect — the daily ritual. One screen. Ten minutes. The Crew's overnight drafts waiting for approval. Swipe through: approve, edit, skip, or mark a contact manual-only. Configurable autonomy per contact and per channel — auto-send trusted categories like birthdays and event confirmations; stage everything judgment-heavy for your review; lock VIPs and family to manual-only. Done in ten minutes over coffee.

No logging. No tagging. No lost relationships. The Crew works the way relationships actually work — in the background, across every channel, reading signals you already generate just by being yourself.

Use of Genspark

Super Agent powers the Crew's real-world research:

Business card enrichment (Scan Crew) — scanned cards trigger Super Agent to open LinkedIn, the company's homepage, Crunchbase, and recent news, then identify mutual connections across the user's graph. A name and title become a full relationship context in under a minute.

Inbound partnership qualification (Inbound Connections Crew) — when a cold partnership email arrives, Super Agent researches the sender and their company — funding stage, product focus, recent posts — so the Crew has real context to qualify the email and draft a reply in the user's voice.

Scribe memory enrichment — voice notes and meeting transcripts pass through Super Agent to extract named people and cross-reference them against the existing Relationship Graph, turning passive recordings into queryable memory.

AI Slides powers the Morning Connect weekly recap — every Monday, Genspark auto-generates a polished deck summarising who the user met, who warmed up, who went cold, and which warm intros are recommended for the week ahead. A shareable artifact the user can forward to their team or investor, generated in seconds.

Image generation powers visual identity — the six Crew characters (Scan, Secret Signals, Inbound Connections, Warm, Cold, Scribe) that appear on the landing page and throughout the dashboard are Genspark-generated, giving each Crew member a distinct visual presence.

Use of Claude

Claude powers the six Crew's core intelligence:

Scan Crew uses Claude vision to extract structured contact data from business card photos — name, title, company, email, phone, LinkedIn, website, plus any handwritten notes on the back. One API call replaces an entire OCR pipeline, and Claude handles messy layouts, multiple languages, and photographic artifacts gracefully.
Secret Signals Crew uses Claude to classify every incoming message across the 11-category Warmth taxonomy — punctuation, greeting style, emoji, response time, length, channel, timing, silence patterns, event proximity, voice behaviour, social signals. Claude doesn't just score the message; it explains its reasoning in a human-readable sentence ("Hi!! + 🔥 + two-minute reply + weekend evening → hot, reactivation"). Every classification has visible reasoning in the UI. No black boxes.
Inbound Connections, Warm, and Cold Crew all use Claude to draft replies in the user's voice — matching tone, reference context (the specific event where they met, past deals they've closed, mutual connections), and proposing concrete next steps. The user's voice samples and organisational context sit in the Profile layer; Claude pulls them into every draft so a message to an ABA contact sounds different from a message to a Claude Code SG builder.
Scribe Crew uses Claude to turn raw voice-note and meeting transcripts into structured memory — extracting named people, commitments, action items, and unresolved questions, then cross-referencing them against the existing Relationship Graph.
Cross-cultural classification leverages Claude's multilingual reasoning — !! means excitement in English but may signal lack of decorum in Japanese business communication, and emoji carry different weight in Chinese. Claude personalises the Warmth classifier per user and per region, which is the technical moat competing tools can't reproduce without multilingual capability at this level.

Claude Code was the build environment for the product itself — scaffold, design system, component architecture, the 11 new dashboard components, fixture data for 25+ contacts, and all server action plumbing. Every file in the repo shows Claude Code's commit history. The disclosure artifact is the build itself.
The product doesn't work without Claude. Strip Claude out and Connect Crew is a UI with no intelligence — no classification, no drafting, no reasoning, no voice-matching. Claude isn't a feature; it's the nervous system of the entire department.
The handoff to Genspark is the product story. Claude reasons and writes; Genspark acts on the web and generates artifacts; Claude writes again with Genspark's findings. Every Crew workflow demonstrates both tools doing what they're best at, with the handoff visible in real time in the dashboard side panel.

What is the deployed URL for this project?

https://connect-crew-xi.vercel.app/