A doorman for your website — but for AI

Per-request authority for the agentic internet.

Software "agents" now visit websites and act for real people — buying things, reading articles, filling in forms. AIdenID checks who each one really is, and decides what it's allowed to do — on every visit. For builders: a four-layer identity check, six clear outcomes, and a signed receipt anyone can verify.

See it work in 20 seconds Request a demo

4 quick checks6 clear answerstamper-proof receiptsverify it yourself

in the background, AI getting checked:allowthrottlequeuesandboxdenypriced

What's quietly leaking

Right now, your best content is leaving — and you can't see it.

↘

When an AI summarizes your article, the reader never arrives — no visit, no subscription, no ad seen.

◆

AI crawlers are now a large and growing share of web traffic — and most of it pays you nothing.

◍

You can't charge for, or even see, access you can't identify. Today, most sites are flying blind.

See it in 20 seconds

Watch one agent get decided — start to finish.

No hunting around. One guided walk: an AI arrives, gets checked, you read its receipt — then watch an unknown one get turned away.

a 20-second guided tour · no clicking around

Don't take our word for it

Ask your own AI about us.

One click drops a ready-made question — comparing us to Cloudflare and Okta — straight into your AI of choice. We're confident enough to send you to a neutral judge.

Ask ChatGPT ↗Ask Claude ↗Ask Perplexity ↗Ask Gemini ↗Ask Grok ↗

This demo has already made thousands of real decisions — here's the running tally.

AI visits checked

every AI visit we've decided on so far

AI visitors turned away

they got a preview — not the real thing

Proof receipts issued

a tamper-proof receipt for every decision

Charged to AI for access

sites can charge AI instead of giving work away free

What can happen

Six ways the doorman can answer.

Like a club doorman, AIdenID gives every AI one of six answers — tap a card to see what each means. Anything but let in means the AI gets only a teaser, never the real content.

200 · full

ALLOW

verified + delegated + in-scope → full resource

tap →

429 · teaser

THROTTLE

allowed but rate-limited → teaser + Retry-After

tap →

202 · teaser

QUEUE

accepted, deferred → teaser + queue position

tap →

200 · teaser

SANDBOX

allowed into a limited view → sandboxed subset

tap →

200 · teaser

DENY

not verified / out-of-scope → teaser only, never full body

tap →

402 · quote

PRICED

pay-to-access → 402 + price quote

tap →

Watch them resolve live →

AIdenID · live enforcement

Watch agents hit the wall — in real time.

👆 Tap any row to open its receipt — that's the AI's verdict, plus proof you can check yourself, right in your browser. Every row is a real decision happening live. The red KILL button cuts off an AI's access on the spot.

ALLOW · 0

THROTTLE · 0

QUEUE · 0

SANDBOX · 0

DENY · 0

PRICED · 0

live request feed/api/aidenid/stream

Tap any request to open its signed decision receipt — and verify it yourself, right here in your browser.

Run this on your traffic How the cascade works →

Use cases · live demos

Six outcomes, every surface.

👆 Pick an AI on the left, then tap the tabs. The same decision plays out across a paywalled article, a data feed, a crawler, and an AI checkout — see what each AI is allowed to do on each one.

Who is requesting?

ShippedGET /blog/aidenid-demoHTTP 200 · ALLOW · body=full

Inside the Agentic Web: Why Your Site Needs Per-Request Identity

AIdenID Research · 9 min read · requested by Googlebot

Half your traffic is no longer human. Crawlers summarize you, shopping agents check out for users, and research bots fan out across your archive — and your access control still assumes a person behind every request.

The fix isn't another CAPTCHA. It's identity. Every agent worth serving is run by an operator that can hold a key — OpenAI, Google, Anthropic — and act under a delegation from a real human principal. AIdenID turns that into a decision you can prove.

A verified, in-scope Googlebot gets the full body. A commercial agent with no delegation gets a price quote. A burst gets throttled. An unknown scraper gets the teaser and nothing more — because the server never serialized the rest.

Gate my surface See the SDK →

How the check works

Four quick questions, asked on every visit.

AIdenID asks these four things, in order. Each AI visit gets one clear answer — plus a receipt that proves it.

Is it really who it says it is?

A cryptographic signature proves the AI comes from a known company like OpenAI or Google — not an impostor. (operator signature)

Did a real person allow it?

It must carry permission from the human it's acting for — like a signed permission slip. (delegation proof)

Is it behaving normally?

A polite visitor is fine; a firehose of thousands of requests a second gets slowed down. (behavioral fingerprint)

Does its company have a good track record?

AI from trusted, well-behaved companies clears easily; sketchy ones don't. (operator reputation)

// one line at the gate — verify, then enforce

import { verifyAgent, enforce } from '@aidenid/sdk'

const decision = await verifyAgent(req, { route: '/blog/:slug', action: 'read', resourceId: slug })

return enforce(decision) // allow→full · priced→402 · else→teaser only

// every decision is a signed, self-verifiable receipt

const ok = ed25519.verify(d.signerPublicKey, d.signature, canonicalString(signingPayload(d)))

For builders: every decision is signed with ed25519 over a canonical encoding — using HTTP Message Signatures (RFC 9421), DPoP (RFC 9449), and OAuth Token Exchange (RFC 8693), with key epochs and replay resistance. Verify any receipt with the public key — no callback to us.

Get the SDK

Every agent · every edge · every tool

Runs where you run. Knows your tools. Decides every agent.

No lock-in: the same decision runs at your edge and recognizes the AI behind today's agents — then clears or blocks each one on its merits.

OOpenAIverified

GGoogleverified

AAnthropicverified

PPerplexitysandbox

MMeta AIthrottled

MMistralverified

BByteDancequeued

CCohereverified

xxAIpriced

AAmazonverified

DDeepSeekdenied

MMicrosoftverified

OOpenAIverified

GGoogleverified

AAnthropicverified

PPerplexitysandbox

MMeta AIthrottled

MMistralverified

BByteDancequeued

CCohereverified

xxAIpriced

AAmazonverified

DDeepSeekdenied

MMicrosoftverified

runs atCloudflare

runs atFastly

runs atNode

runs atPython

runs atVercel Edge

runs atYour origin

knowsOpenAI

knowsAnthropic

knowsLangChain

knowsCrewAI

knowsAutoGPT

knowsCursor

knowsCopilot

knowsTerraform

runs atCloudflare

runs atFastly

runs atNode

runs atPython

runs atVercel Edge

runs atYour origin

knowsOpenAI

knowsAnthropic

knowsLangChain

knowsCrewAI

knowsAutoGPT

knowsCursor

knowsCopilot

knowsTerraform

Built for

Wherever an agent must never be guessed.

The same per-request clearance and signed audit trail — across the highest-stakes domains on the internet.

Defense & Intelligence

Decide exactly which AI can reach sensitive systems — and keep signed proof of every call.

Government

Check every AI that touches citizen data or public services, one request at a time.

Universities & Research

Control which AI can crawl your work, and protect datasets you haven't licensed away.

Financial Services

Only let an AI make a payment if a real person clearly allowed it — and shut it off the instant it misbehaves.

Critical Infrastructure

No unproven AI ever acts on live systems. If in doubt, the answer is no.

Enterprise & SaaS

Let in, slow down, limit, or charge AI for access to your apps and content.

Request a live demo

See it decide on your traffic.

We'll wire AIdenID into a real surface — content, data, APIs, or payments — and show you the live screen, signed receipts, and the KILL switch.