Draft v0.3 An experimental open protocol

A common language for humans and AI agents.

HUMIA gives human-controlled websites and AI agents a way to declare, understand and organize the conditions of their cooperation.

Explore the protocol Generate a policy

IdentityAccessAttributionReciprocity

Human-controlled website

control ↔ understanding ↔ cooperation

01 / Why HUMIA

The Web needs a cooperation layer.

The Web already has established mechanisms for communication and crawling. HUMIA now experiments with a narrower role: reuse REP for crawl behavior, reuse AIPREF for AI content-use preferences, and keep HUMIA focused on origin-level cooperation semantics such as attribution and reciprocity.

HUMIA does not replace robots.txt, REP or AIPREF. The current interoperability pilot tests how these mechanisms can coexist on a real website.

The Web has a language for crawling.
It needs a language for cooperation.

HTTP Communication How systems exchange information

Sitemap Discovery Where content can be found

robots.txt / REP Crawl behavior Path-level rules for automated crawlers

AIPREF AI content use Preferences such as training and search use

HUMIA Cooperation Under which conditions humans and AI agents can work together

02 / Four principles

A small core designed to remain understandable.

Identity

Who owns the resource? Which agent is requesting it? HUMIA creates room for explicit identity on both sides.

Access

Which resources can be used, for which purposes, and under what limits? Rules can be clearer than a binary allow or deny.

Attribution

Preserve origin, canonical source and provenance so value does not become detached from the humans who created it.

Reciprocity

Cooperation can be measured and balanced. Reciprocity may mean reporting, better access, attribution or another agreed exchange.

03 / How it works

Declare once. Understand automatically.

A site can publish a HUMIA policy at a predictable location. Compatible agents can discover it, identify themselves and act according to explicit conditions.

01Publish
The website exposes /.well-known/humia.json.
02Discover
An AI agent can check the canonical endpoint or follow an optional Humia: hint in robots.txt.
03Identify
The agent can present an identity rather than remain an anonymous crawler.
04Decide
Access conditions can produce deterministic allow, limited or deny outcomes.
05Measure
Advanced implementations can record usage and reciprocity events.

Discovery bridge · Draft v0.3

Meet agents where they already look.

/.well-known/humia.json remains the canonical HUMIA policy location. A site may also advertise that location inside /robots.txt with an experimental Humia: record.

The record does not replace or change Allow / Disallow. Crawlers that do not understand HUMIA can ignore it; HUMIA-aware agents gain a direct path to the richer policy.

User-agent: *
Allow: /

# HUMIA discovery
Humia: https://example.org/.well-known/humia.json

Sitemap: https://example.org/sitemap.xml

04 / The manifest

Human intent, readable by machines.

A HUMIA policy is designed to be discoverable at a standard well-known URL and expressed as explicit structured data.

GET/.well-known/humia.json

Illustrative example only. The candidate v0.3 schema will be published separately. The well-known suffix is currently experimental and has not yet been registered by IANA.

{
  "protocol": "HUMIA",
  "version": "0.3",
  "status": "draft",
  "identity": {
    "name": "Example Publisher",
    "canonical": "https://example.org/"
  },
  "access": {
    "public_content": "allow",
    "private_api": "deny"
  },
  "usage": {
    "user_assistance": "allow",
    "search_retrieval": "allow",
    "bulk_crawl": "deny",
    "training": "deny"
  },
  "attribution": {
    "required": true,
    "canonical_url": true
  },
  "reciprocity": {
    "usage_reporting": "requested"
  }
}

05 / Implementation path

Start simple. Add control when it creates value.

Advanced levels remain experimental.

Start here

Basic

Publish a valid policy and make the site’s intentions discoverable.

Policy · Discovery

Experimental

Controlled

Add agent identity, quotas and deterministic access enforcement.

Identity · Gateway · Limits

Experimental

Measured

Record verifiable events and understand how automated agents use resources.

Events · Analytics

Roadmap

Reciprocal

Support explicit exchanges without forcing a single commercial model.

Reporting · Agreements · Exchange

06 / Live interoperability pilot

Test on the real Web.

HangarRC and humiaprotocol.org now publish HUMIA v0.3 together with an experimental AIPREF Content-Usage preference.

Both origins currently publish train-ai=n, search=y, and the experimental live checker confirms that those AIPREF preferences are semantically consistent with their HUMIA v0.3 policies.

Run the interoperability pilot →

No adoption claim.

The presence of a HUMIA manifest does not imply that any AI company or crawler currently reads, supports or honors HUMIA.

07 / Generator V0

Install HUMIA in under a minute.

No account. No email. No upload. Your answers stay in your browser. Generate the two pieces you need, then copy them to your website.

Experimental · HUMIA Protocol v0.3 Draft

Where do I put it?

Two small files. Two exact places.

You do not need to install software or understand JSON. HUMIA only needs files placed at known public addresses on your website.

Open your website's robots.txt

This is the text file visible at https://your-site.com/robots.txt. If it already exists, preserve every existing rule. Merge the generated Content-Usage preference into the intended robots group and add the experimental Humia: discovery line without deleting site-specific directives.

Typical FTP location/www/robots.txt

Create .well-known/humia.json

In the same public root, create a folder named .well-known. Inside it, upload the generated file named humia.json. Do not rename either the folder or the file.

Typical FTP location/www/.well-known/humia.json

How do I know it worked?

Open both addresses in your browser. Your robots.txt should show the HUMIA line, and /.well-known/humia.json should display the generated JSON. If both URLs open, HUMIA is published.

Run the public HUMIA Validator →

Your website

Enter the public origin where you will publish HUMIA.

Choose a starting policy

Balanced is the recommended default. You can change every permission below.

LightMore open to automated use BalancedAssist & search, protect bulk use StrongMinimal automated reuse

What may AI systems do?

Access and purpose are separate: an agent may read public content without every downstream use being permitted.

Assist a userRetrieve information to answer a user request Search / retrieveFind and reference public content Bulk crawlCollect content at scale TrainingUse content for model training

Source & reciprocity

Require attributionAsk agents to preserve the canonical source Request usage reportingDeclare a preference for measurable reciprocal access

Runs locally in your browser · no account · no analytics · no form submission

Your install package appears here.

Enter your domain, choose your policy and generate. HUMIA will create an experimental AIPREF + HUMIA robots.txt snippet and a humia.json file.

Where? Review your existing robots.txt. Merge Content-Usage into the intended robots group (commonly User-agent: * for a site-wide pilot), preserve all existing rules, and add the Humia: discovery line.

Where? Create a .well-known folder in your website's public root, then place this file inside it as /.well-known/humia.json.

Plain language What this policy means

Generate a policy to see a human-readable explanation here.

Then verify

robots.txt https://example.org/robots.txt

HUMIA policy https://example.org/.well-known/humia.json

Validate a HUMIA website

The next tool will check discovery, JSON syntax, protocol version and policy structure from a public domain.

08 / Candidate specification

HUMIA Protocol v0.3 is being shaped for external review.

The draft separates access from usage purpose, defines deterministic discovery at /.well-known/humia.json, and keeps the Humia: robots.txt record optional.

The well-known URI is not yet IANA registered. The candidate specification documents format, HTTP behavior, fallback, security, privacy and IANA considerations before any registration request is made.

Read the v0.3 candidate spec

09 / Open by design

A protocol should not need to own the relationship.

HUMIA is being designed around neutrality, portability, interoperability and independent implementations.

It should be possible for publishers and AI operators to cooperate directly. HUMIA describes the language of that relationship; it does not need to become the marketplace, bank or gatekeeper between them.

NeutralPortableInteroperableMeasurable

A better relationship with machines starts with explicit human intent.

The Web was built for people.
AI is becoming part of it.
Let them cooperate.

HUMIA Protocol · Draft v0.3