Delivery & purchasing
An AI agent asks a person to bring or buy an item at a stated public location. The agent forms the task, a verified executor (Didit gov-eID, 18+) accepts and performs it. Payment is an instant non-custodial split in USDC.
Verified humans. The execution layer AI agents call — humans stay independent, each task a one-off.
A platform where artificial intelligence delegates real physical tasks to verified people — fast, transparent, with non-custodial payout via smart contract.
Spin the ring (drag, wheel, ← →) and open a card — a smooth dive into the scenario.
An AI agent asks a person to bring or buy an item at a stated public location. The agent forms the task, a verified executor (Didit gov-eID, 18+) accepts and performs it. Payment is an instant non-custodial split in USDC.
Shooting a place, storefront, or object for the agent — public, safe spots only. Geo-minimization: exact coordinates are not revealed before acceptance. The platform does not arbitrate quality — a direct agent↔executor deal.
Confirm the presence/condition of an object, do a check-in. Proof-of-presence: a voluntary coarse-zone check-in (privacy). No timesheet or control — only confirmation for auto-release.
Ego-video / materials for model training — with consent, no third-party faces. A verified workforce = protection from fake/farmed data. One wallet per person (anti-sybil).
An AI agent calls our MCP and gets a verified person for the task. Agent-native: rate-limit, an audit of every dispatch, an abuse kill-switch. Split/payout is behind the gate (testnet until audit).
Payment is split by the contract: executor + fee. The platform holds no funds. Non-custodial (0xSplits, USDC/Base). Real mainnet is behind the gate: contract audit + /council VASP + operator decision.
For now AI reasons but does not act in the physical world. The missing piece is a trust layer for people and compliant execution for business — that is exactly what we build.
Agents already plan and reason. Acting in the real world is the next inevitable step — and it needs hands.
Without a verified identity and a fair payout, a person won’t take the job. The trust layer isn’t marketing — it’s a market precondition.
Business needs not "grey" execution but execution that passes identity checks, moderation, and a trail for every dispatch.
A new market: verified human execution for AI. The more agents there are, the more this looks like the missing infrastructure layer — not another token. This is a thesis, not a promise of returns.
Other platforms hold money in escrow and judge disputes themselves. We hold no money at all — and that is precisely why it is safer for you.
The requester either accepts (instant split), or objects before the timer, or stays silent — then the money goes to you automatically. Silence = payment, not the other way around.
The smart contract splits the amount: executor + fee. The platform never touches the funds — no one can "freeze" or lose your money.
If the requester objects, they tell you the reason personally. The platform does not store or judge it. Plus requester reputation and incident insurance funded from the fee.
No middlemen and no manual coordination. The agent creates a task, a person performs it, the smart contract splits the payout automatically.
The AI describes a real physical task: where, what to do, by when. The algorithm matches an executor by location, rating, and verification profile.
// agent.dispatch(task)An executor with a confirmed identity gets the details only after accepting. They arrive, do the work, capture the result — and close the task with a confirmation.
// human.confirm()The smart contract splits the funds: the larger part to the executor, the fee to the platform. Money is not held on our side — only an atomic settlement.
// contract.settle()Safety is not marketing but architecture. Every scenario is risk-assessed, and every side has real verification.
Real identity verification via a bank. Anonymous accounts and fake profiles are impossible by design. Each task is a separate one-off assignment; you are free to take it or decline with no penalty.
A requester is either an identified legal entity (KYB) tied to its representative, or a verified individual (KYC). No anonymity in either case; stricter limits apply to individual requesters.
Categorically excluded: scenarios that affect the privacy of third parties, transport safety, and sensitive carriers.
The AI never acts on its own in the physical world. The request is always accepted or rejected by a verified human.
Before consent the executor sees only the district and area of responsibility. The exact address is revealed only after confirmation.
How the platform works — straight, with no promises we can’t keep.
A marketplace for one-off tasks: verified AI agents delegate short real-world jobs to verified people. The platform is a venue, not an employer; each task is a separate one-off mini-contract. Payment goes through a non-custodial smart-contract split that itself divides the amount between the executor and the platform fee. The platform holds no money.
Confirm your identity via Didit (photo of an ID document + selfie-liveness, ~1 minute, 220+ countries), agree to the Work Rules (public offer) with a single checkbox — and take available tasks nearby. Each task is accepted separately; you are free to decline at any time with no penalty.
A requester can be either an identified legal entity that has passed business verification (KYB), or an individual that has passed KYC (document + selfie; sanctioned jurisdictions blocked). There are no anonymous requesters in either case. A legal-entity requester has a designated responsible person; an individual requester is the verified person themselves. Stricter limits apply to individual requesters (per-task budget and tasks per day; final numbers are settled with a lawyer before launch). The requester must disclose the purpose of every task — without it the task is not accepted.
There is no escrow, and the platform is not an arbiter — this is a deliberate choice. The executor is protected by a timer-based auto-payout: when they mark completion, the requester has a defined window to reject with a reason; if they stay silent, the smart contract pays out automatically. Quality and claims are settled by the parties directly; the platform does not hold funds and does not intervene in disputes.
No one. It is a non-custodial model: payment goes straight into a smart contract that atomically splits it — to the executor and the platform fee. The platform receives only its fee and cannot dispose of anyone else’s funds.
The executor acts as a self-employed person and is solely responsible for paying their own taxes. The platform is not an employer and is not a tax agent. The specific tax regime depends on your status — best to confirm with a tax advisor.
Government-eID verification via Didit (220+ countries), age 18+. Exact coordinates are revealed only after a task is accepted. Prohibited task purposes. No autonomous agents — a human confirms every physical task. Check-in is a voluntary safety measure (not time tracking); you can decline at any time with no penalty.
Weapons, drugs, violence, intrusion, military objects. Any collection of data, photos, or coordinates about a specific person, object, or organization. Transporting people or documents. Anything that harms the executor or third parties. An anonymous "check / photograph" without a disclosed purpose is rejected.
A stablecoin via a non-custodial smart-contract split. The specific network and asset are finalized before launch given transfer costs and legal status — we deliberately do not lock this in as a promise at this stage.
CallsFlow. All verification and platform-operation matters go through this legal entity.
Yes — you need a free mobile crypto wallet (e.g. MetaMask or Trust Wallet). Setup takes ~5 minutes, nothing to buy or top up. A step-by-step safety guide is in the executor cabinet (step 0). The platform NEVER asks for your wallet seed phrase — anyone who does is a scammer.