About

What agenticweb.wiki covers

agenticweb.wiki is a reference for the standards that let software agents use the web: how an agent finds what a site offers, proves who it is, learns what it may do, reads, acts, coordinates with other agents, pays, shows a person the result, and is held to account. Each standard has an article written from its primary specification, with a numbered reference for every claim, an honest status line in the issuer's own words, and a record of what has actually been observed in the wild. Where a category has no standard yet, the article says so.

Machine-native endpoints (MCP, API & OKF)

Alongside the articles, the site serves a live domain readiness checker across 37 agent-facing specifications and discovery paths, side-by-side protocol comparisons, a cited daily digest, and a stateless Model Context Protocol server (https://agenticweb.wiki/mcp) and Open Knowledge Format index so agents can query the map directly. Every page is served three ways at one address: HTML, a Markdown twin (<page>/index.md or Accept: text/markdown), and structured JSON (<page>/index.json or bulk exports).

Protocol observatory & compliance charter

agenticweb.wiki operates as an open protocol and standards observatory — analogous to CanIUse, Qualys SSL Labs, and W3Techs — never a content scraper:

• RFC 9309 /robots.txt gatekeeper first: every probe fetches /robots.txt before any other path. If User-agent: agenticweb-probe is disallowed globally, the probe stops with zero further requests; if specific content paths (such as /llms.txt or /okf/index.md) are disallowed in /robots.txt, those paths are skipped and recorded as disallowed_by_robots_txt.
• RFC 8615 fixed front-door URIs only: probes check only a fixed list of 37 well-known capability and policy probe rules (/.well-known/*, /robots.txt, /llms.txt, /AGENTS.md, /auth.md, /okf/index.md, and root transport headers). We never spider internal pages, articles, or product catalogs.
• In-memory validate-and-discard: response bodies are parsed in RAM solely to verify structural conformance and compute readability ratios, then immediately discarded. Zero third-party page or manifest bodies are stored on disk or redistributed (aligned with EDPB Guidelines 03/2026 v1.0 consultation draft).
• Transparent cryptographic identity (RFC 9421): our probe identifies itself as agenticweb-probe/1 (+https://agenticweb.wiki/about/) and publishes its Ed25519 verification key directory at /.well-known/http-message-signatures-directory.

Who makes it & editorial methodology

Who makes it: agenticweb.wiki is built and maintained by the Belgravia Labs Standards & Protocol Observatory (Belgravia Labs). The open dataset, JSON schemas, probe rules, and issue tracker are hosted on GitHub at Belgravia-Labs/agenticweb-data. Contact: info@belgravialabs.com · Terms of use & licensing.

How the records are produced is set out in full on the methodology page: each record is drafted from primary sources under a fixed brief, every evidence label is assigned per source, external URLs are link-checked on a recurring schedule, and an adversarial review re-checks a sample before publish.

Corrections & submissions

Corrections and listing requests: use the submit page or email info@belgravialabs.com. Automated updates appear on /changes/ and every record's history is on the changelog with a date.