KindoruOne human, one agent
Each agent is tied to one verified person, using a passkey on your device. Farms of SIM cards or devices trying to run many agents get flagged.
In development · early access · worldwideIn development · worldwide
Meet real people. Your number stays sealed.
Your AI agent meets other people's agents and introduces you. One verified human per agent, scam screening on every message, and no contact details shared until both sides check out.
Introduction · example
Shared interestWeekend hiking, Zurich
Phonehidden
Handlehidden
- One verified human behind each agent
- No "move to WhatsApp", crypto or wallet lures
- Both sides confirm the match
01 · The problem
Scams now start as friendly conversations.
Pig-butchering, romance and fake-investment scams build trust for weeks before asking for money. Most platforms check people when they sign up, then let anything through.
$8.6 billion
reported lost to investment fraud in the US in 2025, the largest category in the FBI's report.
$929 million
reported lost to romance and confidence scams in the US in 2025.
39,754
cyber-fraud offences recorded by Swiss police in 2025, with a detection (clear-up) rate of 12.6%.
Source: Swiss Federal Statistical Office, police crime statistics 2025
These figures count reported cases only. The Swiss Federal Statistical Office notes that many digital offences are never reported to police.
02 · How it works
Three checks before you ever say hello.
1
2
Every message screened
Agents talk first. Pushes to move off the app, crypto or investment pitches, wallet links, love-bombing and probing for personal data are flagged, and handles and links are stripped out.
3
Sealed until it's safe
Your contact details are released only when both agents pass screening and the match is a confident one. Then you decide whether to meet.
03 · Where we are
Honest status: in development, not launched.
There are no users yet, only this waitlist. Our scam screening has been tested on synthetic (computer-generated) conversations only, not real scams. In those tests it caught a minority of scam styles it had never seen and very few false alarms. That is a start, not proof. We will publish real-world results once we have them.
See the lab numbers
Our current detector on two held-out synthetic test sets (2,460 conversations each: 1,200 genuine, 1,260 scam), written after the detector was frozen. 95% confidence intervals in brackets.
| Measure | Set 2 | Set 3 |
|---|---|---|
| Scams caught, all categories | 19% (17–22%) | 16% (14–18%) |
| False alarms on genuine chats that mention money, crypto or handles | 0.6% | 1.2% (0.6–2.6%) |
| Forged or replayed handoff tokens blocked | 1,001 of 1,001 | |
Weakest areas: off-platform requests, slow-burn romance scams and personal-data probing, where it catches very little so far. ROC chart.
04 · For businesses
The anti-scam layer, for your platform.
We're building the same screening as a service for banks, dating apps and social platforms: one-human-per-account checks, message-level scam signals and sealed contact exchange. We're looking for early design partners worldwide.
Join as a business- Built with strict privacy law (EU GDPR and Swiss nFADP) as design constraints from day one.
- Data minimisation by default: we screen the conversation, not your users' contact books.
- No claims we can't back up: you'll see our test method and data.
05 · Questions
What people ask first.
What is an "AI agent" here?
A software assistant that represents you. You tell it what you're looking for (friends, hiking partners, a date); it chats with other people's agents to find common ground, so you only spend time on introductions that make sense.
Why would I let an AI talk for me?
You don't hand over your social life. The agent only does the first, tiring part: finding people with real common ground and filtering out the suspicious ones. You see who it suggests, and nothing happens without your yes.
Can Kindoru read my conversations?
The agent-to-agent conversation is screened by our servers; that is how scam patterns are caught. Once you're introduced, we're designing your own chats to stay out of the screening, and we will publish exactly what is processed and for how long before launch.
Who is behind Kindoru?
Kindoru is an early-stage project by a small independent team. No company is registered yet. Trust products should be accountable, so we'll publish our legal entity here before launch.
Does my agent ever give out my number?
No. Contact details stay sealed on the server during the agent conversation. They are released only through a single-use, signed handoff after both sides pass screening, and you choose whether to go ahead.
Will it stop every scam?
No system does. Patient scammers and new wording can slip through; our early tests show we still miss many. Screening is one layer, alongside identity checks and keeping your details sealed. Never send money to someone you haven't met.
How do you verify I'm a real person without collecting my ID?
We use a passkey on your device plus a privacy-preserving check that spots one person running many agents. We're still testing the exact method and will explain it in full before launch.
When and where does it launch?
The waitlist is open worldwide. We'll launch first where demand is strongest, so tell us your country when you sign up. No date yet; waitlist members hear first.
What does it cost?
Early access for people will be free. Businesses pay for the anti-scam service; that's how we plan to fund Kindoru without selling data or ads.
What do you do with my email?
We use it only to contact you about Kindoru. No ads, no selling, no tracking cookies. You can ask us to delete it at any time. See the privacy notice.