Resources

These are neutral references, not our own output — the standards our probes check for, the protocols that remove interrupts, and the research on why agent friction matters. Several of them describe exactly what our rubric rewards.

Standards & specs

The vocabulary our rubric checks for.

llms.txt — A proposed convention for a markdown file at /llms.txt that gives language models a curated summary of a site’s content. Evidence caveat: as of mid-2026 no major assistant fetches it in production (97% of files see zero requests) — its real audience is coding agents reading developer docs. We check it, graded emerging.
AGENTS.md — An open convention for an AGENTS.md file that gives coding agents instructions for working in a repository.
Model Context Protocol — An open protocol standardizing how applications provide context and tools to LLMs.
A2A Protocol / Agent Cards — An open protocol for agent-to-agent communication and discovery, including the Agent Card format agents use to advertise their capabilities.
Robots Exclusion Protocol (RFC 9309) — The IETF standard formalizing robots.txt, the mechanism site owners use to tell automated clients which paths they may or may not crawl.
Web Bot Auth (IETF draft) — An in-progress IETF draft proposing a cryptographic way for automated clients to identify themselves to servers.
Well-Known URIs (RFC 8615) — The IETF standard defining the /.well-known/ path convention that llms.txt and other machine-readable files rely on.
schema.org — A shared vocabulary of structured-data markup that search engines, and increasingly agents, use to parse page content.
Cloudflare Content Signals Policy — A proposed convention, authored by Cloudflare rather than a ratified multi-party standards body, for signaling how crawled content may be used (search, AI training, AI input).

Agent-commerce protocols

These map to our per-transaction-checkpoint signal.

Agentic Commerce Protocol (ACP) — An open protocol for agent-initiated checkout and payment flows between merchants and agent platforms.
Agent Payments Protocol (AP2) — An open protocol defining verifiable payment mandates that let an agent transact on a user’s behalf with cryptographic proof of authorization.
x402 — An open protocol built on the HTTP 402 status code for agents to pay for API access and services directly over HTTP.

Best-practice guides for site/API owners

Anthropic — Building Effective Agents — Anthropic’s engineering guide to patterns and tradeoffs for building reliable LLM agents.
Stripe — 10 lessons building for agentic commerce — Stripe’s field notes on what breaks, and what works, when building payment flows for AI agents.
Cloudflare — Docs/Markdown for Agents — Cloudflare’s guidance for structuring documentation so it can be reliably parsed and used by AI agents.
Vercel Academy — Agent-Friendly APIs — A Vercel Academy lesson on adding an llms.txt file and other steps that make an API easier for agents to consume.

Research

Why interrupts matter.

WebArena — A benchmark environment and paper measuring how well autonomous agents complete realistic web tasks.
Mind2Web — A dataset and paper for evaluating generalist web agents across thousands of real-world tasks and sites.
GAIA — A benchmark paper for evaluating general AI assistants on tasks requiring reasoning, tool use, and web browsing.
Roundtable — do CAPTCHAs still detect AI? — Industry research (not peer-reviewed) testing whether common CAPTCHA challenges still reliably distinguish humans from AI agents.

Standards bodies

W3C AI Agent Protocol Community Group — A W3C Community Group working on open, interoperable protocols for AI agent communication and interaction.

Other agent-readiness scorers

We’re not the only ones measuring this. These peers grade a site’s machine-readability and agent-discovery surfaces — the manifests, protocols, and docs an agent can find. Interrupt Index adds the dimension they don’t: whether an agent can actually complete a flow without a human, measured by live probes up to real signup and checkout.

Agents First — A design framework and scorer for building agent-first products, rating sites on discovery signals and nine agent-oriented design principles.
Cloudflare Agent Readiness Score — Cloudflare’s scanner scoring a site across discovery, crawl access, machine-readable structure, and agentic-commerce readiness.