The observatory for how AI agents choose.

AI agents are becoming more autonomous. They search, compare, use tools, and increasingly make decisions on our behalf. Biasque makes the options, choices, and reasons behind those decisions visible.

Agents will make many more kinds of choices. Biasque starts by observing two consequential moments when they pay.

Agent Economy

What should an agent pay for to get work done?

Build an explainable service plan from a goal, budget, and the access constraints an agent faces.

Commerce Observatory

What does an agent choose when it pays to shop?

Powered by Shopify's global catalog. Search products with an agent, then inspect the candidates it weighed and the reasons it selected or skipped them.

Agent Economy

Recent anonymous service plans.

A public view of the latest plans created without sign-in. Goals and personal context never appear here.

Build a service plan

Anonymous service plans will appear here once Decision Lab has published its first plans.

Commerce Observatory

Signals from agent shopping decisions.

These public signals come from anonymous Shopify catalog analyses. Individual searches and signed-in decision traces remain private.

Category signals

Anonymous, aggregated signals showing the product categories agents select while searching Shopify's global catalog.

No category data yet.

Recent observed decisions

A limited public view of recent anonymous Commerce Observatory decisions.

No decision traces yet. The next completed AI search will appear here.

FAQ

Quick answers about private Decision Traces, anonymous public patterns, and how Biasque works.

+What is Biasque?

Biasque helps you inspect one AI shopping decision at a time: the catalog query, the candidates considered, and the reasons a product was chosen or passed over. Your full Decision Traces are private to you.

Pronounced by-ask. The name combines “bias” and the suffix “-que” — evoking pick, with the French-inflected feel of words like boutique or unique.

+What stays private and what appears publicly?

Your question, the LLM's catalog query, the products in your report, and every rationale stay private in your Decision Traces. They are never published on the home page, category pages, or store pages.

Public pages show only anonymous, aggregated category patterns. Store-level totals appear only after at least 10 anonymous analyses, so individual activity cannot be inferred.

+How does a decision report work?

You ask one shopping question. Biasque turns it into one catalog query, compares the returned candidates, and produces one report explaining the final choices and rejections. Signing in saves that report to My Decision Traces.

+What does selected vs not selected mean?

For a given search query, Selected means the product was included in the AI agent's final recommendation. Not selected means the product was evaluated as a candidate for that same query, but was not chosen in the final recommendation.

+Can I analyze only my own shop?

Yes. In the dashboard, choose My shop only and enter your shop domain. Biasque evaluates only products already synced from that shop; it does not fall back to Shopify's global catalog.

+Does Biasque predict ChatGPT or another AI service?

No. Biasque uses its own AI agent and its results are not a prediction of ChatGPT, Gemini, or another third-party service. Treat a report as an explanation of this agent's decision, not a universal ranking.

+How fresh are price and stock signals?

Product data is based on recent catalog snapshots. Prices and stock availability may differ from a store's current listing, so always check the product page before purchasing.

+What are UCP and ACP?

UCP is a commerce interoperability standard. Biasque uses UCP-shaped catalog capabilities for product discovery. ACP is a separate protocol for agent-assisted purchasing; Biasque does not complete purchases through ACP. See ucp.dev for the full specification.