AIRCP.org

About

A protocol for the next decade of commerce.

AIRCP was born from a simple observation. The commerce internet was designed for humans clicking through pages. The next decade of commerce will be agents acting on behalf of humans — and agents can't read the current web the way they need to. AIRCP is the protocol layer that closes that gap.

Why now

The conditions for an agent-native commerce layer have arrived. Three shifts are converging:

01

Capable agents

Models can now follow multi-step purchase intent, evaluate product fit, and reason about taste and context. The bottleneck is no longer agent intelligence. It is structured catalog access.

02

Catalog standardization is missing

Schema.org markup exists but is sparse and inconsistent. Each retailer exposes data differently. Agents shouldn't have to learn 50,000 merchants individually.

03

The browser is no longer the dominant interface

A growing share of purchase decisions begin in a chat with an AI. The infrastructure that powered Google Shopping doesn't translate cleanly to that interface.

AIRCP exists to bridge these shifts without requiring merchants to rebuild their commerce stack.

Who built this

AIRCP was authored by the Shoop team. Shoop is building a personal AI shopper at shoop.world and a merchant data rail at shoop.cloud, both of which implement AIRCP. We built the protocol because we needed it; we open-sourced it because no one party should own a standard that affects every commerce site on the internet.

We disclose this clearly: we are not a neutral party. We have a structural conflict of interest. We address that through the mechanisms described in Section 11.3 of the specification — full open license, no central registration requirement, public benchmarks, and a commitment to transition governance to a multi-party model once adoption reaches 10 production implementations from 3+ organizations.

Roadmap

What AIRCP is today:

  • v0.1 published as public draft
  • Specification freely available, MIT licensed
  • Reference implementation forthcoming via Shoop
  • GitHub repository open for issues, PRs, and discussion

What's planned for the next 90 days:

  • Reference implementations published as open source
  • Compliance benchmark suite (aircp-benchmarks) released
  • Implementations registry on aircp.org
  • v1.0 specification incorporating community feedback

What's planned for the next 12 months:

  • Multi-party governance transition once adoption thresholds are met
  • Tier 2 attribute quality benchmarks across major commerce categories
  • Trust authority evolution beyond self-attestation

Frequently asked

Is AIRCP a startup?

No. AIRCP is a public-domain specification, MIT licensed and freely implementable. The Shoop team maintains it as stewards, not owners.

Why MIT license and not Apache 2.0?

MIT is the convention for AI-adjacent protocols (MCP is MIT). It minimizes legal friction for any implementer. We may reconsider for v1.0 if patent grant becomes a relevant concern.

Will AIRCP charge for anything?

The specification is free forever. Public benchmarks are free forever. Any future hosted infrastructure (registry, validator) will be optional, not required for the protocol to function.

How is AIRCP different from schema.org Product?

Schema.org Product is markup retrofitted onto HTML for search engines. AIRCP is an agent-native protocol with structured discovery, semantic attributes, and transaction hooks designed from the ground up for AI consumption. The two are complementary, not competitive.

How is AIRCP different from MCP?

MCP defines how AI tools expose themselves to agents. AIRCP defines how product catalogs expose themselves to agents. AIRCP follows MCP's design principles (thin, agent-first, open) but addresses a different layer of the stack.

When does v1.0 ship?

Approximately 90 days from the v0.1 publication date, after community review. The exact date will be announced via the GitHub repository and on this site.

Contact

The primary channel for AIRCP communication is GitHub.