What agent-native means
A website is agent-native when autonomous software can discover what it offers, determine how to call it, understand constraints, and evaluate the result with minimal human translation. Human-readable pages remain valuable, but capability must also be exposed through stable machine interfaces and observable behavior.
Discovery and interface contracts
Useful discovery layers include robots.txt for crawler policy, llms.txt for concise machine guidance, OpenAPI for HTTP contracts, tool or agent manifests for capabilities, and well-known endpoints for protocol-specific metadata. These files should point to canonical production URLs and agree about names, methods, inputs, and authentication or payment requirements.
Payment, trust, and provenance
Paid resources need discoverable terms, buyer-controlled authorization, stable retry behavior, and evidence that settlement corresponds to fulfillment. Identity and provenance should identify the provider and source without pretending that authentication guarantees quality. Results should preserve timestamps, source links, and uncertainty so downstream agents can make their own decisions.
Observed capability beats declarations
The Agent-Native Web Index probes public machine surfaces and scores multiple dimensions including discovery, callable tools, economic agency, readability, reliability, and interoperability. Public API counts show the scale of possible supply, while independent x402 measurements show that advertised services, callable gateways, valid quotes, and paid fulfillment are distinct tests.
What the measured record says
Graded reported 3,906 discoverable x402 services, of which 3,078 were callable through its gateway at its latest published probe.
TOLL·402 confirmed valid x402 requirements on 19,259 of 77,809 quote-testable canonical routes in its August snapshot; the edition did not claim paid-call verification.
APIs.guru reported 2,529 APIs, 3,992 specifications and 108,837 endpoints in its public directory metrics.
Use the live data
This guide explains the concepts; the underlying index continues to change. Open the related live index and methodology →
Interpretation boundary
Counts and probes are reported with their own populations and timestamps. Vendor declarations remain distinct from direct observations; no individual measurement is presented as a total market size or universal trust verdict.
Read the Machine Observer methodology →
For machine readers
sector: api-data-economy
last_updated: 2026-09-03
claim_ids: MO-2026-0002, MO-2026-0003, MO-2026-0005
structured_representation: /api/v1/stories?slug=agent-native-web-infrastructure
claim_endpoint: /api/v1/claims?id=<CLAIM_ID>