anml-client

ANML 1.0 client library for Python

---

A modern, async-first Python client for the ANML (Agent Negotiation Markup Language) 1.0 specification. Built for AI agents that need to discover, negotiate, and interact with machine-first web services.

Features

• Async-first — Built on httpx for modern async HTTP
• Type-safe — Full Pydantic v2 models for all ANML types
• Secure — defusedxml for safe XML parsing, HTTPS enforcement, SRI verification
• Discovery — Well-known, Link header, and HTML link discovery
• Disclosure — Consent management and disclosure evaluation
• Flow navigation — Step-by-step flow traversal utilities
• Pagination — Async iteration over nav links
• Builder pattern — Fluent client configuration

Installation

pip install anml-client

Quick Start

import asyncio
from anml_client import AnmlClient, AllowListTrustPolicy

async def main():
    # Build a client with trust policy
    client = (
        AnmlClient.builder()
        .base_url("https://api.example.com")
        .trust_policy(AllowListTrustPolicy(["example.com"]))
        .build()
    )

    async with client:
        # Discover ANML endpoint
        anml_url = await client.discover("https://example.com")
        if anml_url:
            # Fetch and parse the ANML document
            doc = await client.fetch_url(anml_url)

            # Navigate flows
            from anml_client import FlowNavigator
            nav = FlowNavigator(doc)
            print(f"Current step: {nav.current_step}")
            print(f"Flow complete: {nav.is_complete}")

            # Execute actions
            response = await client.execute_action(
                doc, "get-profile", params={"user_id": "123"}
            )
            print(f"Response: {response.status_code}")

            # Paginate through results
            async for page in client.paginate(doc):
                print(f"Next page: {page.head.title}")

asyncio.run(main())

Parsing ANML Documents

from anml_client import parse_anml_xml, parse_anml_json

# Parse XML
doc = parse_anml_xml(xml_string)

# Parse JSON
doc = parse_anml_json(json_string)

# Auto-detect from content type
from anml_client import parse_anml
doc = parse_anml(content, "application/anml+xml")

Disclosure & Consent

from anml_client import ConsentStore, evaluate_disclosure, AllowAllTrustPolicy

store = ConsentStore()
store.grant("email", purpose="communication")

# Evaluate disclosures in a document
from anml_client import evaluate_all_disclosures
results = evaluate_all_disclosures(doc, store, AllowAllTrustPolicy())

SRI Verification

from anml_client import compute_sri, verify_sri

# Compute integrity hash
sri = compute_sri(data, "sha256")

# Verify data integrity
is_valid = verify_sri(data, "sha256-...")

API Overview

Module	Description
anml_client.client	Main AnmlClient with builder pattern
anml_client.types	Pydantic models for all ANML 1.0 types
anml_client.parser	XML and JSON parsing
anml_client.config	Client configuration and trust policies
anml_client.discovery	ANML endpoint discovery
anml_client.disclosure	Consent management and disclosure evaluation
anml_client.action	Action execution with parameter binding
anml_client.flow	Flow navigation utilities
anml_client.knowledge	Knowledge exchange helpers
anml_client.integrity	SRI verification (SHA-256/384/512)
anml_client.pagination	Async pagination over nav links
anml_client.middleware	HTTP middleware chain
anml_client.errors	Exception hierarchy

Links

• ANML Foundation
• ANML Specification
• GitHub Repository

License

ISC