Stated intent beats behavioral profiles

Twenty years of ad tech built an expensive machine for one purpose: guessing what a person wants from what they did last month. Cookies, device graphs, lookalike audiences — an entire industry of inference, plus an entire regulatory regime (GDPR, FTC consent decrees) generated by its side effects.

Agent conversations make the guessing machine obsolete. When someone asks a flight tool for "TLV to BKK in November," intent isn't inferred — it's stated. The most valuable targeting signal in advertising history is just... there, in the tool call.

So that's all we take

Our SDK sends six allowlisted context keys: tool, category, query, route, locale, country. Values truncated to 200 characters, filtered client-side before a request is even built. No transcripts. No user IDs. No cross-session anything — the schema has nowhere to put them.

This isn't privacy as a compliance checkbox. It's privacy as a better product:

• For users — no profile exists to leak, subpoena, or creep on.
• For publishers — nothing sensitive transits your infrastructure; there's nothing to audit because there's nothing there.
• For advertisers — stated intent converts better than inferred intent, because it isn't a guess.

The old machine needed to know who you are to guess what you want. The agent economy tells us what the moment wants and nothing else. Match the moment, settle on conversions, forget everything. That's the whole trick — and it's a moat precisely because incumbents can't adopt it without burning down their own surveillance assets.