The US thesis

US e-commerce is a $1.1 trillion annual market (Census Bureau). Services — rides, home services, food delivery, rentals, healthcare appointments — is another multi- trillion-dollar layer. Agents can already browse, summarise, and recommend; what they cannot yet do at scale is transact. The missing piece is a transactional protocol with consent, payment, dispute, and audit baked in.

By 2030, a plausible fraction of US consumer transactions will start as “hey ChatGPT/Claude/Gemini, book me a…” A Manhattan doctor visit, an Austin home repair, a Chicago restaurant reservation, a Miami airport transfer — all agent- mediated. The question is whether the plumbing that supports those transactions is a handful of proprietary walled gardens (Amazon agents, Google agents, OpenAI agents) or an open protocol that anyone can implement.

Why this is a US regulatory question, not just a tech one

FTC: Section 5 (unfair/deceptive), the Made in USA Rule, endorsement rules, and the CARS Rule all apply to agent-mediated transactions in different ways. Who is the “endorser” when an agent recommends a product?
CFPB: Reg E, EFTA, the Remittance Rule, authorised-push-payment scams — all designed assuming human-initiated transactions. Agent-initiated payment requires rule update or clarification.
State attorneys general have been active on consumer protection, from California to New York to Texas. Agent transactions that go wrong will land on their desks.
DOJ antitrust + FTC: agent marketplaces risk tipping into monopoly if one platform's agent defaults to its own supply. Open protocols help pre-empt this.
Section 230: when an agent makes a recommendation, is the platform a publisher? Case law is not ready.
US state AI laws: Colorado AI Act, California ADMT rules, NYC AEDT already touch high-stakes uses; agent commerce will attract the next wave.

What we think 2030 looks like in the US

A Gen-Z consumer in Austin, Texas, spends forty seconds on their phone: “find me a licensed plumber for a leaking bathroom, book for tomorrow morning, budget $250.” The agent queries multiple marketplaces, negotiates, locks a slot, escrows payment, confirms via text. The agent also consults the user's privacy policy (via GeraMind-like context vault) and a compliance layer (GeraCompliance) to ensure the transaction meets the user's rules. The plumber gets the job at a fair rate; the consumer gets the task done; the platform's take rate is transparent and bounded.

Why GeraNexus from the US

The Gera Services portfolio already operates 28 consumer and prosumer products across healthcare, home services, commerce, food, transport, rentals, and more. That is a rare supply-side footprint to anchor a transactional protocol against. We can eat our own dog food for the first year, then open the spec publicly. US-compliant reference implementations will target CFPB Reg E and FTC deceptive-practices precedent from day one.

Roadmap — US context

Q3 2026: public spec v0.1, reference implementation across five US-available Gera verticals (GeraClinic, GeraHome, GeraRide, GeraMarket, GeraEats).
Q4 2026–Q1 2027: private pilots with two non-Gera US marketplaces, FTC-and-CFPB- ready compliance matrices published.
2027: spec v1.0, open governance, state-AG engagement.
2028–2030: cross-protocol interoperability with MCP, A2A, and emerging standards.

Competitors and peers

Anthropic MCP: read-heavy protocol. GeraNexus layers transactional semantics on top.
OpenAI Apps / Operator: proprietary; not an open spec.
Google agents and Project Astra: proprietary; strongly tied to Google's stack.
Agora, Skyfire, Nanda: each attempting parts of this; none pairs with actual live supply-side liquidity at Gera's scale.

US cross-links

See the US-specific thinking on related 2030-infra products: GeraMind (personal context vaults), GeraVoice (voice-native commerce), GeraWitness (human oversight).

US sources

US Census Bureau — Quarterly E-commerce Retail Sales
FTC — Unfair or Deceptive Acts or Practices (Section 5)
CFPB — Regulation E; 2024 rulemaking on larger participants
Colorado AI Act (SB 24-205) and California ADMT rulemaking