Machine-First Architecture launched in March 2026 describing four abstractions: identity, structure, content, and interaction. The protocols that implement those abstractions have been arriving ever since. This page maps them to the pillars.
Each entry includes a status label. Production means it is shipping in real products today. Recommendation and Candidate Recommendation are W3C standards track designations. Early Preview is in browser or platform beta. Draft is specification work in progress. Emerging is a community proposal with traction but no formal standards body. These labels are accurate as of April 2026 and will age.
The point of the map is not to memorise every protocol. It is to see that every pillar of the framework now has real infrastructure backing it. The theoretical part is over.
1.Identity
Machine-verifiable identity is moving from schema markup into cryptographic proof. The direction of travel: identity a machine can verify, not just recognise.
Verifiable Credentials (VC)
RecommendationW3C
Cryptographically secure, privacy-respecting, machine-verifiable credentials. VC v2.0 reached Recommendation status in 2025, establishing the baseline for portable identity claims on the web.
W3C
Globally unique identifiers that can be verified without a central registrar. DIDs v1.1 is at Candidate Recommendation, specifying the syntax and resolution process that makes identity portable across platforms.
Trusted Agent Protocol (TAP)
ProductionVisa
Adds cryptographic proof-of-identity to agent-initiated transactions using signed HTTP messages carrying intent, verified user identity, and payment details. Co-developed with Cloudflare. Partners include Stripe, Shopify, Microsoft.
Know Your Agent (KYA)
EmergingTrulioo
An identity, trust, and audit framework specifically for agents, built with PayOS. Introduces a Digital Agent Passport describing who built the agent, who it represents, and what permissions it carries. Source is a vendor whitepaper describing Trulioo's own product.
2.Structure
Structure is the oldest pillar and the one with the most mature tooling. Schema.org plus JSON-LD remains the baseline. The shift is in how that structured data gets consumed once the graph is in place.
Schema.org + JSON-LD
RecommendationSchema.org Community
The lingua franca for structured data on the web. JSON-LD is the fastest-growing format, now appearing on more pages than any other structured data syntax.
GraphRAG
ProductionMicrosoft Research
Extracts knowledge graphs from raw content, builds community hierarchies, and uses that structure for retrieval-augmented generation. Demonstrates what explicit relationship architecture produces at scale.
llms.txt
EmergingJeremy Howard / Answer.AI
A Markdown file at site root offering a curated, machine-readable map of key content. Adoption is growing but hard to measure precisely. No major AI platform has formally confirmed inference-time consumption, but the convention is spreading.
3.Content
The content pillar is fusing with the licensing layer. Publishers are moving from "content as pages" to "content as licensed, versioned, structured assets" with machine-readable usage terms.
Really Simple Licensing (RSL)
EmergingRSL Standard
A framework for publishers to declare how AI systems may use their content and at what price. Adopted by Arena Group, BuzzFeed, USA Today Co, and Vox Media.
IAB Tech Lab
Machine-readable payment instructions for AI content licensing. Establishes structured compensation models spanning pay-per-crawl, aggregate licensing, and attribution-based pay-per-use.
4.Interaction
The fourth pillar attracted the most scepticism at launch because its infrastructure did not exist yet. It exists now. The three-protocol stack for agentic interaction (agent-to-tool, agent-to-agent, agent-to-website) is shipping in production.
Model Context Protocol (MCP)
ProductionAnthropic / Linux Foundation AAIF
The de facto standard for agent-to-tool communication. Adopted by Anthropic, OpenAI, Google, Microsoft, and AWS. Donated to the Agentic AI Foundation under the Linux Foundation in December 2025.
Agent-to-Agent Protocol (A2A)
ProductionDefines how agents discover and coordinate with each other. Agents publish capability descriptions at /.well-known/agent-card.json. Backed by 150+ organisations.
WebMCP
Early PreviewGoogle / Microsoft / W3C
Lets websites declare capabilities as structured tools through a navigator.modelContext API. W3C Community Group specification. Shipping in Chrome 146 early preview.
Agentic Commerce Protocol (ACP)
ProductionOpenAI / Stripe
Live in ChatGPT since September 2025. Merchants expose ACP endpoints; ChatGPT creates checkout sessions using Shared Payment Tokens scoped to specific merchants, amounts, and time windows.
Universal Commerce Protocol (UCP)
ProductionAnnounced at NRF 2026. Defines discovery through purchase through post-purchase as structured protocol surfaces in Google Search AI Mode and Gemini. Partners: Shopify, Etsy, Target, Walmart, Visa, Mastercard.
IETF
PKCE, DPoP, and Token Exchange (RFC 8693) to bind tokens to agents and swap broad human tokens for narrow task-scoped agent tokens. Continuous access evaluation for real-time revocation.
Linux Foundation
Governance body for MCP, Goose, and AGENTS.md. Co-founded by Anthropic, OpenAI, and Block. Platinum members include AWS, Google, Microsoft, Bloomberg, and Cloudflare.