huggingface.co agentic web readiness

A Score 57 (coverage 30 + breadth 12 + conformance 15). 6 standards observed across 3 of 9 steps; 5 manifests present, 1 error, 6 warnings. Checked 2026-10-18 22:06:01 UTC. This grades the agent-facing files published at huggingface.co, not the organisation's products.

Badge for a README: ![agentic web grade](https://agenticweb.wiki/badge/huggingface.co.svg) · JSON · Markdown · Re-check now

Coverage by step

01 DiscoverAI Catalog well-known file, RFC 9727 02 Identifynothing observed 03 PermitOAuth Protected Resource and Authorization Server Metadata (as used for agent authorization)04 Readnothing observed 05 Act 06 Coordinatenothing observed 07 Paynothing observed 08 Presentnothing observed 09 Accountnothing observed

Manifests

Manifest	Result	Findings
Agentic Resource Discovery `/.well-known/ard.json`	absent HTTP 401
AI Catalog well-known file `/.well-known/ai-catalog.json`	valid HTTP 200	entries[1] has no description; registries index on it
MCP Server Cards (SEP-2127) `/mcp/server-card`	valid HTTP 200	remotes[0].supportedProtocolVersions missing; clients use it to pick a version before connecting
Agent2Agent Protocol `/.well-known/agent-card.json`	absent HTTP 401
llms.txt `/llms.txt`	absent HTTP 404
AGENTS.md `/AGENTS.md`	absent HTTP 404
api-catalog: A Well-Known URI and Link Relation to Help Discovery of APIs `/.well-known/api-catalog`	valid HTTP 200
AI Crawler User-Agent Tokens in robots.txt `/robots.txt`	valid HTTP 200	no AI crawler token named (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, ...); the policy for them is whatever User-agent: * says (allowed) no Agentmap: line (optional convention pointing agents at /.well-known/ard.json) no Content-Signal: line (optional; states ai-train / search / ai-input preferences)
security.txt (RFC 9116) `/.well-known/security.txt`	errors HTTP 200	Contact: security@huggingface.co is not a mailto:, https: or tel: URI Canonical: missing
auth.md `/auth.md`	absent HTTP 404

What an agent can call

MCP endpoint https://huggingface.co/mcp (streamable-http) — from /mcp/server-card
API description https://huggingface.co/.well-known/openapi.json (application/json)
robots.txt AI policy: unlisted crawlers allowed
OAuth protected resource https://huggingface.co/mcp, authorization servers: https://huggingface.co

Standards observed

AI Catalog well-known file https://huggingface.co/.well-known/ai-catalog.json
api-catalog: A Well-Known URI and Link Relation to Help Discovery of APIs https://huggingface.co/.well-known/api-catalog
OAuth Protected Resource and Authorization Server Metadata (as used for agent authorization) https://huggingface.co/.well-known/oauth-protected-resource
MCP Server Cards (SEP-2127)
Model Context Protocol
AI Crawler User-Agent Tokens in robots.txt

Cheapest next steps

llms.txt: one Markdown index file; the most widely read agent-facing file
Agentic Resource Discovery: a catalog at /.well-known/ard.json lets agents discover your APIs, MCP servers and data in one fetch
AGENTS.md: instructions for coding agents working on or with the site

How the grade is made

+10 per lifecycle step with a standard observed, counting manifests that validate or that the probe rule for the standard recognises (a valid MCP server card counts for MCP; max 90); +3 per observed standard beyond the first two (max 15); conformance +15 (+10 with one manifest) minus 5 per agent-facing manifest with errors, never below 0 (security.txt is reported, not scored). A+ >= 60, A >= 45, B >= 30, C >= 20, D >= 10. Errors are specification violations; warnings are best-practice gaps. Only public paths are fetched, without credentials, so standards that need a login (skipped in the probe) are not counted either way. The same report is the validate_manifest tool on the MCP server.

JSON · Markdown · probe rules