Question 1

What is Sage at sagepays.xyz?

Accepted Answer

An AI agent that turns one product URL and one budget into paid user-testing missions, then pays human testers in USDC for verified evidence. It settles on two mainnet rails — GOAT Network for public receipts and Starknet for private payouts — and is registered on-chain as ERC-8004 agent #79.

Question 2

How do I get paid to test products on Sage?

Accepted Answer

Open the Sage marketplace, pick a mission, do the short task it describes, and write what you actually saw. There is no application and no interview. Sage checks your account against what it observed exploring the product itself and pays in USDC.

Question 3

Does Sage actually pay, or is it a claim?

Accepted Answer

Every payout is an on-chain transaction on GOAT Network or Starknet and publishes a receipt page anyone can open and verify. The marketplace lists recent payouts with links to those transactions, so the total paid is read from settled transfers rather than typed in.

Question 4

Who pays the testers?

Accepted Answer

The agent does, autonomously, out of a campaign vault the founder funded. Each payout is a transaction with a public receipt, and the founder never reviews a submission by hand.

Question 5

What stops the AI agent from overspending?

Accepted Answer

The agent proposes, a smart contract disposes. Each campaign has an on-chain vault that derives the exact reward, enforces the per-mission cap and completion limits, and rejects anything outside them. No model output can move money by itself.

Question 6

Do I have to review submissions?

Accepted Answer

No, and that is the product. Sage judges each report against its own first-hand observations and pays the genuine ones. Work it cannot verify is held for a human decision rather than guessed at.

Question 7

Can I write my submission in my own words or my own language?

Accepted Answer

Yes. Sage matches your account against phrases from inside the product rather than requiring particular wording, so a genuine account written in your own voice or another language still verifies. One thing to keep: write the product's own words for what you saw on screen, such as button and heading labels, because those are what Sage matches against. Copying the mission description back scores nothing.

Question 8

Do I need a crypto wallet to launch a campaign?

Accepted Answer

No. The Telegram bot @sagedeputybot runs the whole loop walletless: it creates a server wallet bound to a spending policy and funds the campaign from it, so you never connect anything or leave the chat. A browser wallet on sagepays.xyz works too if you prefer one.

Question 9

What does a founder give Sage to start?

Accepted Answer

A product URL, what you want proven, who should test it, and a budget in USDC. Sage browses the product itself and designs the missions, so there is nothing else to fill in.

Question 10

What if my product is behind a login or a Connect Wallet screen?

Accepted Answer

Sage inspects everything it can reach, and when a wall stops it, it reads the product's own documentation to understand what is behind it, then designs missions a tester who has their own account can genuinely perform. It will never ask a tester for credentials, a seed phrase or a private key.

Question 11

How long does an inspection take?

Accepted Answer

Usually about ninety seconds to a plan you can read, and you watch the browsing happen on screen while it works, page by page.

Question 12

What does Sage not do?

Accepted Answer

It is not a bug bounty, a generic agent platform, or a chatbot. It turns one product and one budget into paid, verified testing, and it is judged on whether the payouts hold up.

Trading calls with a record nobody can fake.

Three ways to use it.

The Oracle

Signals

Your own agent

Written before. Checked after. Never deleted.