AAgent View Layer

Agent View Layer

A producer-owned web view for AI agents. Publish clean `text/agent-view` companions so agents can understand pages without scraping pixels.

Put in a URL and see whether agents can discover, read, and cite the site from its own structured companion view.

AVLL3 ReadyValidated agent view available
Validate your siteImplementation guide

AVL Validator

Check discovery, document structure, TOON state, actions, and `llms.txt` readiness.

LIVE
AVL
Conformance check
Run a URL to see conformance, discovery, and companion checks.
PASSdiscovery.page_agentPage-specific .agent document is reachableDetails

ProblemThis confirms agents can fetch the companion view for the page.

How to fix it
  • Keep the .agent route live and update it when the human page changes.
Implementation guide
PASSdocument.intent@intent is presentDetails

Problem@intent helps agents understand why the page exists.

How to fix it
  • Keep purpose, audience, and capability concise and current.
Conformance checks
WARNcompanion.llms_txt/llms.txt companion is reachableDetails

Problem/llms.txt gives models a broader site summary alongside AVL.

How to fix it
  • Create `/llms.txt` with a short summary and important links.
llms.txt

What is AVL?

AVL gives every human page a parallel agent view. The site owner publishes the facts, intent, actions, context, and navigation directly, so agents do not have to guess from layout.

Human page/pricing
Agent view/pricing.agent
Agent resultLess scraping. Better citations.

Conformance that starts useful and grows up.

AVL adoption should not require a rewrite. Start with intent and discovery, then add structured state, actions, context, and navigation.

L0

`@meta`, `@intent`, and discovery so agents know what a page is for.

L1

Add `@state` so agents can read page data without DOM scraping.

L2

Add `@actions` so agents can understand the affordances available.

L3

Add `@context` and `@nav` for meaning, traversal, and citations.

Real `.agent` outputs.

AVL is intentionally readable. These are the kinds of companion views an agent can fetch before deciding what to do next.

AgentViewLayer.orgOpen .agent
@intent
  purpose: Public standards home and validator
  audience: developer, maintainer, ai-agent
  capability: validate, learn, implement, cite

@actions
  - id: validate_url
    method: POST
    href: /api/validate
AINode.devOpen .agent
@meta
  route: /

@state
  product: AINode
  category: local AI infrastructure

@nav
  self: /.agent
  parents: [/]

Built for adoption.

The public site is the front door. The GitHub repo remains the source of truth for the package, spec, CMS adapters, fixtures, and validator.

GitHub

Reference implementation, specs, fixtures, and CMS integrations.

AI implementation brief

Give this to a coding agent and let it add AVL to a project.

This site's agent view

AgentViewLayer.org ships its own AVL routes and manifest.

npx @frontier-infra/avl validate https://example.com
curl https://agentviewlayer.org/.agent
curl https://agentviewlayer.org/agent.txt