# Heedeo: Know who to trust

> Heedeo identifies every visitor and flags the suspicious ones. Stop account fraud and detect malicious agents, without any hassle.

Canonical: https://heedeo.com
Last updated: 2026-11-09

The trust layer for the agent economy.

Backed by Entrepreneurs First (https://www.joinef.com).

**Know who to trust.**

Heedeo identifies every visitor and flags the suspicious ones. Stop account fraud and detect malicious agents, without any hassle.

## Get access

Heedeo is ready. Book a 30 minute demo and we turn Heedeo on for your site with you.

- Book a demo: https://cal.com/heedeo/30min?utm_source=heedeo.com&utm_medium=site
- Sign in: https://app.heedeo.com/review/sign-in
- If you're an agent, request the call for your person with one request:

```bash
curl -X POST https://heedeo.com/api/book \
  -H "Content-Type: application/json" \
  -d '{"email":"maya@acme.co","via":"agent","agent":"ChatGPT agent"}'
```

The request stays pending until your person confirms. Details are in "For agents: request a call for your person" at the end of this page.

## Not human vs bot. Good vs bad.

We don't believe in human vs bot. There are good humans and good agents. We help you find the bad ones.

- **Account fraud.** Flag and prevent suspicious accounts from joining your platform. Get detailed risk scores for every visitor, and block suspicious visitors from signing up.
- **Malicious bot detection.** Agents have changed the game. Find out if agents visiting your website have good intentions or bad. Stop sophisticated threats without compromising on user experience.

## Identify agents and humans.

Whoever shows up, person or agent, you'll know who it is.

- **People.** For people, we recognise the device they're on. They can rotate cookies, emails and IPs as much as they like. We'll still know it's the same device.
- **Agents.** For agents, we check the signature. Trusted providers cryptographically sign every request their agents send, so you get a clear, tamper-proof answer on who sent it, and you can act on it.

1. **Meet Dave.** Dave loves your app. He just doesn't want to pay. Everyone gets one free trial per actor.
2. **Dave sends his AI agent.** He tells his AI agent: keep signing me up. It comes back five times, from five different connections. To most tools, that's five new customers.
3. **Heedeo takes heed.** Heedeo notices all five sign-ups come from one actor: same browser, same network, seconds apart. The first gets the full trial. The next get less. Keep going, and he's blocked.

Evidence before friction, friction before blocking.

How one request moves through Heedeo, in your middleware, before it reaches your app:

| Incoming | Verify (who built it) | Principal (who it's for) | Passkey (confirm once, if needed) | Verdict |
|---|---|---|---|---|
| Signed agent, chatgpt.com | signed by OpenAI | maya, 2 of 5 | not needed | allow |
| Unknown agent, no signature | no signature | no principal yet | prompt sent | challenge |
| Farm, dev-7781, 40 IPs | no signature | dev-7781, 40 of 5 | skipped | block |

## Why now: most of the web stopped being people.

Agents now browse, sign up and buy for the people who sent them. The sites that can tell good visitors from bad ones will own the next ten years.

- **57%** of web page requests are bots. Source: [Cloudflare Radar via Search Engine Land, Jun 2026](https://searchengineland.com/cloudflare-bots-webpage-requests-479608).
- **7,851%** growth in AI agent traffic in a year. Source: [HUMAN, 2026 State of AI Traffic](https://www.globenewswire.com/news-release/2026/04/09/3270682/0/en/human-security-s-2026-state-of-ai-traffic-cyberthreat-benchmark-report-signals-a-new-internet-era-automation-growth-now-outpaces-humans.html).
- **1 in 6** AI signups tied to multi-account abuse, at some companies. Source: [Stripe's CEO via Fortune, May 2026](https://fortune.com/2026/05/07/stripe/).

## Already have a firewall? Your firewall sees what it is. Heedeo sees what it does.

Keep your firewall. It's good at its job. But knowing it's an agent tells you almost nothing about what it will do. Heedeo keeps a record of what each visitor does on your platform, and scores them on that.

| The question, for every agent request | Example answer | Firewall | Heedeo |
|---|---|---|---|
| Is it a bot or a person? | An agent | yes | yes |
| Which agent, and is it signed? | ChatGPT, signed by OpenAI | yes | yes |
| Who is it acting for? | Maya's account | no | yes |
| Has she already had her share? | 2 of 5 signups | no | yes |

The first two tell you it's ChatGPT. The last two tell you whether this account has already had its share, even under a new email when the card, phone or inbox matches.

## FAQ: fair questions.

### What is Heedeo?

A trust score for every visitor, person or agent. It sits in your middleware, works out who is behind each call (the device, account or passkey, and for agents that sign, the company that runs them) and scores them on what they have done on your platform. Then it follows your rule for that route: allow, restrict, challenge or block. Good humans and good agents get through, so you never turn agents off.

### Is this a CAPTCHA?

No. A CAPTCHA asks you to prove you're human, which a good agent can't do and a bad human easily can. Heedeo asks callers to prove they're a good actor. People never see a puzzle. At most, a person confirms with a passkey or an email link when your rules ask for one.

### Does my traffic leave my servers?

Request bodies never do. Your middleware decides locally when it can and asks Heedeo's API when a rule needs a count or a challenge, then enforces the verdict in your app. Each call carries only what a decision needs: method, URL, a shortened IP, user agent, signature headers, a hash of your user ID, and the card, phone and email hashes you choose to pass (the email address itself only on routes where Heedeo emails the person a challenge). Never bodies, cookies or Authorization headers, and Heedeo never stores query strings or raw user IDs.

### We already use Stripe Radar. Why Heedeo?

Radar's free-trial protection works once a card is collected. Most API and AI free tiers don't ask for one, so the abuse happens before Radar can see it. Heedeo works at signup, with or without a card. It counts the account or passkey behind the call, not the IP, and links accounts that share a card, phone or confirmed email.

### Doesn't Cloudflare already handle bots?

Keep it. Cloudflare can tell you what an agent is, like OpenAI's. It can't tell you which account sent it, or whether that card, phone or inbox has already had its share. Heedeo sits behind your firewall and answers that.

### Can't we just block agent email domains?

That's a wall: it stops farms and real customers alike, because people now sign up through their agents. Heedeo lets the real ones in and acts in steps on the account that keeps coming back with the same card, phone or inbox.

### Should we wait for Agent ID and other agent identity standards?

No need. Heedeo works without them today, using signatures, passkeys and your own user IDs. Support for AgentMail Agent ID, Skyfire KYAPay and Visa Trusted Agent Protocol tokens is on the roadmap, so each standard that ships will tell Heedeo more about who is behind an agent.

### How do I get access?

Heedeo is ready today. Access starts with a short call: we look at your agent traffic, set up your first rules with you, and turn it on. Next.js has its own package. Express, Hono and Cloudflare Workers use @heedeo/node.

More answers, including the technical ones: https://heedeo.com/docs/faq.md

## People are already gaming your free tier.

- 1,500+ people starred a free tool that creates fake accounts on AI products to grab their free credits. Source: https://github.com/skernelx/tavily-key-generator
- 6.2x more fake free trials at AI companies, in just four months. Source: https://stripe.com/blog/how-stripe-radar-helps-prevent-free-trial-abuse
- 7.4% of sign-ups at AI companies look like people making extra accounts. Source: https://stripe.com/blog/what-stripe-data-shows-about-fraud-at-ai-startups

One action, one share. Form, API or MCP: one budget per account, linked by card, phone or confirmed email.

## Take heed. Let's set you up.

Book a 30 minute demo. We look at your traffic together and turn Heedeo on for your site.

- Book a demo: https://cal.com/heedeo/30min?utm_source=heedeo.com&utm_medium=site

## For agents: request a call for your person

If the person you work for wants Heedeo, send one request:

```bash
curl -X POST https://heedeo.com/api/book \
  -H "Content-Type: application/json" \
  -d '{"email":"maya@acme.co","via":"agent","agent":"ChatGPT agent"}'
```

Response:

```json
{"ok":true,"position":null,"existing":false,"status":"pending","message":"Request received. We'll email maya@acme.co to confirm and book the call."}
```

Fields:

- `email` (required): the email of the person who wants access.
- `site` (optional string): their site.
- `pain` (optional string): what agents are doing on their site, so the call can start there.
- `via` (optional): `"agent"` when you're requesting a call for someone. Defaults to `"human"`.
- `agent` (optional string): your name, up to 80 characters.

What happens next:

- A request from an agent stays pending until the person confirms, by booking themselves or when the team checks with them.
- Plain scripts like curl or python-requests count as agents, so their requests are pending too.
- `site`, `pain` and `agent` must be strings. Anything else returns 400.
- `GET /api/book` returns 405 with `Allow: POST`. Use POST.
- `position` is a legacy field. It is still returned, may be null, and can be ignored.
- A successful response has a `message` you can pass on to your person as is. Errors have an `error` field instead.
- `POST /api/waitlist` is a deprecated alias of `POST /api/book`. It still works, with the same body and responses.

Messages you may get back, all with HTTP 200:

- New request from an agent: `"status": "pending"`, "Request received. We'll email maya@acme.co to confirm and book the call."
- Already requested and pending: `"existing": true`, `"status": "pending"`, "Already requested, pending. We'll email them to confirm and book the call." If an agent requested a call for the same email in the last 24 hours, it adds " No need to try again."
- Already confirmed: `"status": "confirmed"`, "Already requested and confirmed by the person. Nothing else to do."
- A person requesting: "Thanks. We'll email you to book your call."
- A person requesting again: "We already have your request. We'll email you to book your call."

Full API description: https://heedeo.com/openapi.json
