Be generous to humans.
Stingy to farms.
One call tells your backend how many free credits this signup should get. When the same person signs up again, the second trial is smaller.
The product
- Verify with a credit grantOne call returns a decision and the credit share to give.
- Repeat accountsLink the next signup to accounts already seen.
- Email intelligenceAliases, disposable and brand-new domains, as one signal.
- Observe modeRead the farm report without changing anything: would have stopped N, est. $.
- SDK and quickstartBrowser token, one backend call. Node, Next.js, Python.
Also in the box
- Payment guard: hold paid credits until money settles.
- Clawback: signed events when fraud evidence lands later.
- API key guard: spot stolen or resold keys.
- MCP client guard: check DCR registrations and client farms.
- AI agents: verify agent identity before applying policy.
- Agent budgets: cap volume per verified operator.
- AI crawlers: block training crawlers, allow AI search.
- SMS pumping: check SMS destinations before sending.
- Promo abuse: code guessing, leaked codes, repeat redemptions.
- Referral abuse: self-referrals and linked referral rings.
- Invite abuse: keep phishing out of your invites.
- Abuse network: opt in to hashed reports from other owners.
- Session hijack: a stolen cookie used from another machine.
- Login protection: new devices on returning accounts.
- Step-up: proof-of-work or email check for challenged users.
- Bot detection: automation signals behind the signup.
- Usage abuse: account pools and fast credit burn.
- Rules: turn signals into an ordered policy.
- Appeals: give a wrong block a way out.
Start in observe mode.
Change credits later.
Observe mode allows traffic and returns the proposed decision and grant. Your backend keeps control of account creation and credit balances.