Concept build. The Parallax beta is planned for Black Friday, 27 November 2026.
Read →Before you or your agent check out, witness agents in other places and profiles quote the same item at the same store, at the same moment, each with proof that the price came from the retailer's site.
Check my priceSame item, many vantage points, one truth.
Quoted from other places and profiles. If your price is the odd one out, you see the gap before you pay.
Check my priceThe browser extension, or the MCP tool for agents, sends the retailer, item, zip code and timestamp to the network.
Volunteers run headless browsers from their own locations and profiles. Each returns the price with a zkTLS proof that the page came from the retailer.
Your price, the median of the witnesses who answered, and the gap between the two. A confidence level separates likely profiling from noise.
A public board ranks retailers by measured price spread, live through Black Friday on 27 November and Cyber Monday on 30 November. Quote hashes and timestamps are committed onchain, so no number can be backfilled. Every check produces a receipt card you can share.
The check costs a few cents in USDC, paid over x402 on Base. An agent with one fixed profile is the easiest shopper to recognise, so it is the one that should ask most.
Add the browser extension. At checkout it sends the retailer, item, zip code and timestamp, then shows your price next to the witness median before you pay.
Give your agent the MCP tool. It pays for the check itself over x402 and gets back the same comparison.
// MCP tool call parallax_check({ retailer: "<store>", item: "<item>", zip: "<zip>", at: "<timestamp>" }) → yours 48.20 · median 44.00 · gap +9.5% · conf <level>
Checks open with the Black Friday beta.
Checks are paid in USDC: 80% goes to the witnesses who answered and 20% to the treasury, which buys back $PLAX. Shoppers never need to hold the token.
This holiday season, AI agents do a real share of the shopping, and an agent with one fixed profile is the easiest shopper to recognise and price.
Maryland's limits on surveillance pricing took effect.
Requires disclosure of algorithmic personalised prices.
Black Friday. The beta and the spread board go live.
Cyber Monday. The board keeps updating.
In a study by Consumer Reports and Groundwork, volunteers checked the same Instacart items at the same store and time. That was a one-off study. Parallax runs the same check whenever a buyer asks.
A gap means you were priced differently. It does not say why. Here is what is known, and what is still open.
Stock limits, coupons, shipping, login state and A/B tests all cause gaps. Each result carries a confidence level, and the method is published. When too few witnesses can load the item in time, there is no result at all.
A retailer could block witness traffic or show witnesses clean prices. Both are detectable, and either would be news in itself.
Retailers' terms forbid automated visits, and witnesses could receive cease-and-desist letters. This is not resolved.
Each quote carries a zkTLS proof that the page came from the retailer. Witnesses also post a bond, forfeited if a spot re-check exposes a fabricated quote.
zkTLS proofs work on many large retailers but break site by site. The beta covers a limited set of major retailers, not every store.
Attention drops after the holidays. The public board and the new state rules on personalised pricing give people a reason to keep checking.
Shoppers pay in USDC. $PLAX rewards the people who run witnesses, and it is not the product.
