The human check everyone can call
Deepfakes broke every trust signal you rely on. One API call answers the only question that matters now: is there a verified human on the other end?
One query, one answer
// POST /v1/verify { "handle": "@alan", "tier": "live", "require": ["human", "identity"] } // response { "human": true, "proofs": ["human", "reach", "identity"], "score": 72, "freshness": "2026-01-12T09:41:00Z", "revoked": false }
Integrate once
One REST API. Add "Verify with hmn" wherever fake humans cost you money: signups, listings, chats, calls, transactions.
Choose your bar
Require named proofs ("Human + Identity"), a minimum score ("70 and above"), or both. Different actions can demand different bars.
Query per session
Each check returns proofs, score, freshness, and revocation status. Pay per query, priced by assurance tier and band.
The same check, asked of agents
Agents are starting to act for people everywhere. The moment one reaches you on someone's behalf, you face the same question in reverse: is there a verified human behind this agent, and what is it allowed to do? The endpoint you already call to verify a human answers it. Same rail, no new integration.
// POST /v1/verify (agent-provenance · in pilot) { "agent": "aide:alan", "acting_for": "@alan", "require": ["human_behind_agent", "scopes"] } // response { "human_behind_agent": true, "scopes": { "acts": ["dispute", "negotiate"], "reads": ["subscriptions", "disputes"] }, "freshness": "2026-01-12T09:41:00Z", "revoked": false }
No anonymous agents. Every agent carries a real human behind it.
A verified human stands behind the agent, and the scopes the delegation permits. Not that the agent's output is correct or the person's intent is good. It attests real, not good, exactly like the human check.
The reverse check stays pairwise-pseudonymous. You confirm a verified human plus this specific scoped agent, with no identifier that follows the person across services.
This answers the human-behind-the-agent question the emerging agent-payment standards (Google AP2, Visa Trusted Agent Protocol, Mastercard agentic tokens) assume but leave open. hmn does not move money or issue mandates. It verifies the human an agent answers to, and what that agent may do.
Agent-provenance is in pilot. Live query docs open to pilot partners after the security review completes.
Three assurance tiers
"Verified" means different things at different stakes. You pick the tier per action.
| Tier | Proves | Mechanism | Typical use |
|---|---|---|---|
| Present | The verified human's enrolled device | Device signs a challenge | Gating DMs, marketplace chat |
| Live | The human is at the device now | Signature gated by biometric unlock | Transactions, account changes |
| Challenge | A live human on this session, right now | Fresh camera liveness, session-bound, replay-proof | Video calls. The $25M wire scenario. |
Every verification is challenge-response, timestamped, and pairwise-pseudonymous: platforms cannot collude to track a person across services. The same ladder sets an agent's authority: routine actions run on standing delegation (Present), larger ones re-engage the human's biometric (Live), high-value or anomalous ones demand fresh liveness (Challenge).
The hmn Score
The Badge is binary: a verified human, or not. The Score is a second layer on top, measuring identity strength through proof groups the person adds voluntarily.
| Proof group | Signals | Verified via | Status |
|---|---|---|---|
| Proof of Human | Liveness, uniqueness, device attestation | Camera liveness, dedupe, secure-enclave key | mandatory |
| Proof of Reach | Mobile number | WhatsApp verification, telco signals | mandatory |
| Magic link, domain-age signal | optional | ||
| Proof of Identity | Government ID | Aadhaar OTP, DigiLocker, NFC passport | booster |
| Proof of Residence | Utility bill, address | Bill APIs where available | booster |
| Proof of Means | Account ownership | Penny-drop, Account Aggregator, open banking | booster |
| Capacity attestation | Consented statement or income read | booster |
Every verified parameter strengthens the score. You set the bar per action: named proofs, a minimum score, or both. Exact weights and score bands are being finalised with pilot partners.
What you get. What we never share.
hmn sells attestations, never information. Documents live in the person's Vault. Our Notary reads once, attests, and forgets.
Proofs held, score, band, assurance tier, freshness timestamp, revocation status. Everything you need to trust the session. For an agent check: whether a verified human stands behind it and the scopes it carries, never who that human is.
Names, ID numbers, documents, statements, biometrics, or anything that lets you track a person across other platforms. Nothing that makes you a data liability.
Honest assurance
Nobody can promise 100%, and anyone who does is selling to you badly. What the cryptography makes certain: a valid signature is proof the enrolled credential authorized this session. The residual risk lives at the edges, and we attack every edge. The agent claim carries the same honesty. We attest the human behind the agent and its scopes, never the agent's judgment.
- Adversarial liveness. Enrollment and Challenge-tier checks designed to beat real-time deepfakes.
- Velocity anomaly detection. One human verifying 500 listings an hour is a rented credential. Auto-flagged.
- Instant revocation. Stolen device or compromised credential dies network-wide in one call.
- Periodic re-attestation. Proofs age; freshness is part of every response.
Assurance level, freshness, revocation. Delivered by API, with SLAs and liability we underwrite. Carrying that risk is not our weakness. It is the moat.
Put a verified human behind every session.
Pilot the API in a fraud-heavy flow: signups, listings, high-value calls. We'll bring the humans.
platforms@hmn.ai