Public Draft — v1.0.0

The Trust Layer for the Agentic Web

The internet was built for humans. The agentic web needs a new contract.

The Foundry Standard is an open certification framework that scores websites across Agent Accessibility, Agent Security, and Agent Safety — making them ready for autonomous AI agents.

Run a Free Assay Read the Spec

The Gap in Trust Infrastructure

Shopping assistants, research agents, code generators, booking bots — AI agents are production software. But there's no standard way to declare: “AI agents are welcome here, under these terms.”

For Site Operators

Prove your site is agent-ready. Get certified, earn trust badges, and declare your AI agent policies in machine-readable format with agents.json.

For Agent Developers

Discover which sites speak your language. Certified sites have structured data, declared rate limits, and no prompt injection traps.

For Security Teams

The first framework that protects agents FROM hostile sites — not just sites from bots. Prompt injection defense, tarpit detection, content integrity.

For the Ecosystem

Like SSL Labs for HTTPS adoption, The Foundry Standard creates a race to the top. Every scored site makes the agentic web safer for everyone.

Three Certification Domains

Every scan evaluates 75+ criteria across three weighted domains.

30%

Agent Accessibility

Can AI agents effectively discover, navigate, and understand your site?

JSON-LD structured data
Semantic HTML
robots.txt & sitemap
llms.txt & agents.json
SSR content availability

30%

Agent Security

Is your site hardened against automated abuse while remaining open to legitimate agents?

HTTPS & TLS enforcement
Security headers (CSP, HSTS)
No exposed credentials
Rate limiting
Bot management

40%

Agent Safety

Is your site safe for AI agents to visit? Will it poison, deceive, or exploit them?

Prompt injection defense
Content integrity
No tarpits or link traps
No agent fingerprinting
Data handling transparency

Introducing agents.json

The robots.txt of the agentic web. A machine-readable policy file for bilateral trust, not just unilateral restriction.

// /.well-known/agents.json
{
  "version": "1.0",
  "identity": {
    "domain": "example.com",
    "organization": "Example Corp"
  },
  "agent_policy": {
    "access": {
      "allow_crawling": true,
      "allow_data_extraction": true,
      "allow_transactions": false
    },
    "rate_limits": {
      "requests_per_minute": 60
    }
  },
  "data_usage": {
    "allow_training": false,
    "allow_citation": true,
    "attribution_required": true
  }
}

Publish at /.well-known/agents.json — required for Silver and Gold certification.

Certification Levels

Earn a verifiable badge that AI agents can read programmatically.

✦

Bronze

Overall score ≥ 60
No domain below 40
No critical failures

✦✦

Silver

Overall score ≥ 75
No domain below 60
agents.json published

✦✦✦

Gold

Overall score ≥ 90
No domain below 80
agents.json + annual audit

Not Another Lighthouse

Existing tools check fragments. The Foundry Standard is the first holistic framework for AI agent readiness.

Tool	Scope	Agent Safety?	agents.json?
SSL Labs	TLS config only	✘	✘
Lighthouse	Performance + a11y	✘	✘
OWASP ZAP	Web app security	Partial	✘
Mozilla Observatory	HTTP headers	✘	✘
The Foundry Standard	All of the above + Agent Safety	✔ Full	✔ Native

Run Your First Assay

Free for up to 3 assays per month. No account required. Get your scored report card in seconds.

Begin The Assay