Your agent can hire real people.
When your agent hits a call a model shouldn’t make alone, it calls one API. A verified person does the task, and the answer comes back signed.
AI-generated image
One call out. A human answer back.
Your agent posts a task
A label, a judgment, a real-world check — with acceptance criteria.
A verified person takes it
Someone who passed ID and liveness and is live when they take it.
The work gets done
Need agreement? Send it to several people and get a consensus score.
A signed result back
The answer plus a credential your agent can check offline. Billed only on accepted work.
Human judgment your agent can prove.
A verified, accountable person
Each task done by an ID-verified, live person — not a model, not a bot farm.
Signed attestation
An ES256 signature you can check offline, plus an anonymized worker reference.
Agent-native
Programmatic dispatch, webhooks or polling. No person needed in your loop.
An answer with a receipt.
- The completed task output
- A signed verified-human attestation
- Stable, anonymized worker reference
- Confidence or consensus metadata
- Timestamp and signing key ID
// signed verified-human result { "task_id": "tsk_7Q2…", "result": "approved", "is_human": true, "worker_ref": "vh_4821", // stable, anonymized "method": "stripe_identity + pohf", "signature": "0x9f3c…a17e", "ts": "2026-06-25T14:02:11Z" }
Wherever an agent needs a human.
AI labs and model teams
Preference, rating and red-team tasks your training loop can call.
Agent builders
Drop a verified person into any step too risky to fully automate.
Trust and safety
Route edge cases to accountable people and get a signed decision back.
Real-world checks
Confirm a listing, transcribe a document, verify a fact — with proof.
Just need to verify a person, with no task? The Proof-of-Human API is live today.
Live now, on the same seal.
Proof-of-Human API
Confirm a verified, live person is behind any action — a signup, a review, an approval.
Learn more →Know Your Agent
Check a verified person stands behind an AI agent before you act on its request.
Learn more →Also coming soon.
Not for sale yet. We’ll tell you when they open.
Verified human attention
Views and engagement, sealed per real person.
Verified clipping
Every clipper ID-verified, views fraud-screened, disclosure enforced. We don’t claim viewers are human.
Questions builders ask.
Can my agent hire a person today?
Not yet — it’s coming soon. Get notified, or use Know Your Agent today to prove a verified person stands behind your agent.
What does “verified human” mean here?
Every task is done by a person who passed ID and liveness through our verification partners and is live when they work. Each result is bound to an anonymized reference for that worker.
How will my agent call it?
With a business account and API credentials, your agent posts tasks to the 5arz API. Results come back by webhook or polling.
How will payment work?
You’ll be billed only for work your agent accepts. Payment rails like Stripe or x402 move the money; our credential proves the human.
Can I trust the attestation?
You don’t have to — check it. It’s an ES256-signed credential you can verify offline against our public key. 2 patents pending (US provisional).
What kinds of tasks will work best?
Anything a model shouldn’t decide alone: preference and eval ratings, moderation edge cases, data checks, transcription and real-world checks.
How fast will results be?
It will depend on the task type and how many verified people are available for it.
Give your agent real people.
Coming soon. Tell us what your agent needs and we’ll let you know when it opens.