Agentic Notation
Markup Language
A machine-first markup language for agent-to-agent and agent-to-service communication over the internet. ANML describes content, intent, and interaction patterns optimized for machine interpretation.
Introducing ANML /ˈæn.ɪ.məl/
ANML, pronounced “animal,” is a machine-first markup language designed for the agentic web. Where HTML serves humans, ANML serves the autonomous agents that are becoming the primary consumers of web content.
We call ANML files “duckuments.”
A duck moves along the water with effortless grace. Straightforward, easily understood, calm on the surface. But beneath the waterline there is a lot going on: structured interactions, disclosure constraints, workflow state, knowledge exchange, and behavioral guidance all working together.
An ANML duckument looks simple to read and author, but it carries everything an autonomous agent needs to understand a service, respect privacy boundaries, navigate multi-step workflows, and act on behalf of its user with minimal inference over unstructured markup.
Machine-First
Structured for direct interpretation by autonomous agents, minimizing inference cost and maximizing determinism.
Dual Serialization
XML for human authoring and review. JSON for programmatic generation and agent-pipeline consumption. Both normative.
Privacy by Design
Agents control disclosure. Services may request information but constraints govern what can be shared.
What an ANML duckument looks like
This example shows a concert ticket service teaching an agent about a 4-week early bird deal, venue policies, accessibility info, media assets with descriptions, and a multi-step purchase flow — all in a single document.
What an ANML document defines
Content
Structured, semantic information for machine interpretation
Interactions
Operations bound to HTTP methods, endpoints, and parameters
Knowledge
Bidirectional information exchange between services and agents
Constraints
Disclosure, consent, and authorization rules agents must evaluate
State
Metadata identifying the current phase of a multi-step interaction
Persona
Advisory behavioral and tonal guidance for agent responses