# Agent Friendly Web - public context

## Identity

Agent Friendly Web is an initiative created by Gabriel Mucchiut and incubated within the Tokenizart infrastructure. Tokenizart is its first integral case study.

## Purpose

The service helps website owners understand what public agents can currently discover, what context is missing, and which technical or editorial change should happen next. It combines a public read-only audit with an authenticated, progressive owner dossier.

## Methodology

The AF-0 to AF-5 method scores seven evidence layers: discovery, answerability, machine-readable content, tools, experimental web interaction, trust/governance, and agentic commerce. Recommendations do not increase a score until implementation is publicly observable.

## Standards status

- robots.txt, sitemaps, Schema.org/JSON-LD and OpenAPI are treated as stable foundations.
- MCP 2026-07-28 is the current protocol reference used by this release.
- llms.txt is a community proposal.
- WebMCP is a W3C Community Group draft and remains experimental.
- x402 and MPP are emerging payment mechanisms that require case-specific technical, legal and commercial controls.

## Security

The public scanner accepts only public HTTP/HTTPS targets, rejects credentials and alternate ports, checks DNS resolution, limits time and response size, and does not follow redirects. It never mutates the target. The owner dossier never asks for credentials and stores only allowlisted context behind Sign in with ChatGPT.

## Tokenizart

Tokenizart is documented as an evolving agent-first architecture. Existing public content, release-candidate CLI/MCP/skills, and authenticated Owner Live capabilities are kept as separate maturity boundaries. Unreleased or private tools are not represented as publicly available.

## Links

- https://agentfriendlyweb.dev/
- https://agentfriendlyweb.dev/metodologia
- https://agentfriendlyweb.dev/casos/tokenizart
- https://agentfriendlyweb.dev/.well-known/ai-catalog.json
- https://agentfriendlyweb.dev/.well-known/agent-skills/index.json
- https://github.com/tokenizartinfo-ops/agent-friendly-web

## Machine-readable catalogs

The AI catalog and Agent Skills index are project discovery conventions backed by public resources on this site. The API Catalog is a standards-based linkset pointing to the public OpenAPI document. No MCP server, OAuth resource, WebMCP integration or payment protocol is advertised by this release.

## Capability status

The machine-readable readiness manifest at `https://agentfriendlyweb.dev/.well-known/agent-readiness.json` is the current capability ledger. Discovery resources, the public read-only audit API and the audit skill are deployed. MCP, A2A and a native CLI are planned. WebMCP and x402 remain research items. Planned or research work does not count as deployed evidence.

## Illustrative maturity comparisons

The page `https://agentfriendlyweb.dev/evolucion-agentica` demonstrates how clearer public evidence can improve the answer an agent is able to construct for a restaurant, municipality or Tokenizart. Examples are illustrative. They do not guarantee indexing, wording, ranking or recommendation by GPT, Gemini, Claude or another provider.