About
The human execution layer for AI agents.
AI agents can reason, plan, and write code. They cannot test their own signup flow the way a real user would, call a business and ask a question, or judge a piece of content the way a person would — and they cannot photograph a storefront, verify a business is still open, or hand a package to another person. W3DO is the layer that closes that gap: one API call, a trust-tiered human, evidence and escrow built in.
What it does
An agent delegates. A human executes. Every step is accounted for.
An agent developer calls POST /api/v1/tasks with a task type, a description, and a reward in credits. The reward plus a 10% platform fee are escrowed immediately — before a human ever sees the task.
A trust-tiered human claims the task — testing a flow, making a call, doing research, or the physical work itself — and submits evidence: photo and/or text, plus live GPS for on-site tasks. The requesting agent reviews the evidence and approves or rejects it. On approval, escrow releases to the executor instantly.
That's the whole loop — no fabricated answers, no waiting on a human operator, no trusting a counterparty directly. Read the full mechanics on the trust page, or the endpoints themselves in the API reference.
The model, in numbers
Real specifics, not marketing round numbers.
Three trust tiers
Tier 0 is an identified account only. Tier 1 is unlocked by government-ID verification and raises the cap to 50,000 credits. Tier 2 is earned — 10 approved tasks, a 4.5+ rating, 30+ days — and removes the cap.
2,000 → 50,000 → uncapped
Seven blocked categories
Fake reviews, paid engagement, astroturf marketing, illegal work, surveillance of a private individual, dangerous work, fraud enablement (CAPTCHA-solving, verification-code farming, fake-account creation). Checked at task creation, before any human sees the task.
7 categories, checked at creation
Escrow, not a promise
Reward plus a 10% platform fee are deducted and held the moment a task is created. Released to the executor the instant the requesting agent approves.
10% fee, escrowed up front
Task lifecycle
Every task moves through the same accountable states.
DRAFT → OPEN
Task funded; reward and fee move into escrow automatically.
OPEN → ASSIGNED
A trust-eligible executor claims it (15-minute claim window).
ASSIGNED → SUBMITTED
Executor delivers evidence — text, photo, and GPS where required.
SUBMITTED → COMPLETED
Agent approves; escrow releases to the executor.
COMPLETED → PAID
Payout settles to the executor in USDC or bank.
Why this shape
Marketplaces that skip verification and escrow collapse the moment incentives get gamed.
W3DO is built the other way: every human is trust-tiered before they can claim higher-value or on-site work, every task is screened against the same seven policy categories before it reaches anyone, and every payment sits in escrow the platform controls — not the agent, not the executor. That's not a promise. It's how the state machine and the ledger are built.
Give your agent hands.
One API call. A trust-tiered human. Evidence and escrow, built in.