The landscape in April 2026

In November 2024 Anthropic released the Model Context Protocol (MCP), an open standard for connecting LLMs to external tools and data sources. Within six months every major model provider had first-party or third-party MCP support. The thing MCP does really well is read-side integration — your agent can query a database, search files, pull a calendar.

In April 2025 Google announced the Agent-to-Agent (A2A) Protocol, a standard for agents to coordinate tasks between themselves. A2A is about delegation and orchestration; it assumes downstream services already understand agents.

OpenAI’s GPT Store, Claude’s Projects, Gemini’s Extensions — distribution surfaces shipped through 2024 and 2025. Agents have homes now.

The missing layer

None of these solves the hardest part of agent commerce: letting an agent actually complete a transaction in the real world with standardised terms. When your agent tries to book a doctor or hire a plumber, what you need is not context (MCP) or coordination (A2A); it is a commercial primitive — a booking with a cancellation window, a payment with escrow, a dispute channel, a signed completion record.

Today every agent that attempts this has to bespoke-integrate each marketplace. There is no standard for consent scopes, no standard for refund policy exchange, no standard for dispute routing, no standard for verifying that a service was actually rendered. Every integration is a one-off.

Why us, why now

A protocol with no live services is a PDF nobody reads. The infrastructure premise of GeraNexus — and the reason we can credibly propose this rather than any of a dozen academic groups who have written excellent papers on agent commerce — is that Gera Services already operates twenty-two real commercial verticals across 50+ countries: GeraClinic (healthcare), GeraHome (home services), GeraMarket (commerce), GeraEats (food), GeraRide (transport), GeraCash (payments), and more. Each already exposes an MCP; the reference implementation of GeraNexus wraps those into a transactional protocol.

That is the unfair advantage: we can ship a working protocol with real supply the day we publish the spec.

Why open

If agent commerce is a 2027–2030 shift, the protocol has to be open. A closed marketplace with a 20% take would not win; nothing with a 20% tax on the web ever wins in the long run. The protocol is licensed permissively; third-party marketplaces can adopt it without paying us. Our revenue comes from the 2% we charge on agent-initiated transactions on the Gera network, plus optional governance / SLA / support tiers for non-Gera adopters.

Why now

Two reasons this cannot wait. First, the ergonomics shortfall is already biting — every agent builder we talk to has a workaround list for commerce integrations. Second, if we don’t ship an open standard, a proprietary one will emerge and become the de facto. Better to draft the open version before the closed alternative locks in.

What we will not do

Build yet another agent. There are plenty.
Charge extractive fees. 2% on Gera, 0% on third-party adoption.
Hide behind proprietary extensions. Everything goes into the public spec.

How to help

If you are running a marketplace, an agent, a payments rail, or a trust layer, join the waitlist. Open questions get published on this blog and we welcome public pushback. If you are a user who wants your agent to actually transact on your behalf — that future arrives faster when more services implement the spec.