# Vendor facets — Cloudflare. The one CDN/edge vendor here that ships things the Kin Score reads:
# managed robots.txt with Content Signals (consent_identity), a security.txt served from the
# provider's own zone (well_known_published), and an open-source OAuth provider library for Workers
# that serves RFC 8414 / RFC 9728 discovery documents and DCR. The lift it can cause directly is
# small (3 AR points plus about 10 composite points). The large OAuth/MCP lift only happens if the
# provider builds and deploys the server itself. Caching, TLS, DDoS, WAF, rate limiting,
# pay-per-crawl and x402 earn nothing at 0.22.0.
vendor: cloudflare
name: Cloudflare
website: https://www.cloudflare.com
areas:
- mcp-hosting
- cdn
- hosting
registry_keys:
- cloudflare
- cloudflare-pages
rubric_schema_version: 0.22.0
generated: '2026-09-25'
features_refreshed: '2026-09-25'
basis: measured
summary: >-
  Cloudflare lifts a provider's score directly in two places. Its managed robots.txt adds Content Signals,
  which earns the agent-readiness consent_identity dimension. Its free security.txt is served from the
  provider's own zone at /.well-known/security.txt, which earns well_known_published and supports the
  security-disclosure check. The bigger agent-readiness surface (protected-resource metadata, dynamic
  client registration, served OAuth discovery, a remote MCP server) comes from the open-source Workers
  OAuth provider library and the Agents SDK. That lift is conditional: it reaches the score only when
  the provider builds the server, puts it in front of its own API on its own domain, and the probe finds
  it. Being on Cloudflare does not earn it. The edge itself earns nothing: caching, TLS, DDoS, WAF, rate-limiting
  rules, API Shield enforcement, pay-per-crawl and x402 paywalls. Markdown for Agents is real, deployed
  and unscored at 0.22.0.
features:
- id: content-signals-robots
  name: Managed robots.txt with Content Signals Policy
  description: >-
    Cloudflare prepends a managed block to the zone's robots.txt, or creates the file, with Content-Signal
    directives (search, ai-input, ai-train) and Disallow rules for known AI crawlers.
  source: https://developers.cloudflare.com/bots/additional-configurations/managed-robots-txt/
  tier: all
- id: security-txt
  name: security.txt generator and edge serving
  description: >-
    The provider fills in security.txt fields (Contact, Expires, Policy, and so on) in the dashboard or
    API, and Cloudflare builds the file at the RFC 9116 /.well-known/security.txt path on the provider's
    zone.
  source: https://blog.cloudflare.com/security-txt/
  tier: all
- id: markdown-for-agents
  name: Markdown for Agents (Accept text/markdown negotiation)
  description: >-
    When a client sends Accept text/markdown, the edge converts the HTML response to Markdown. It adds
    x-markdown-tokens, and adds a permissive content-signal header when the origin sets none.
  source: https://developers.cloudflare.com/fundamentals/reference/markdown-for-agents/
  tier: paid
- id: ai-crawl-control
  name: AI Crawl Control
  description: >-
    Per-crawler allow/block rules, AI-crawler traffic analytics and robots.txt violation tracking for
    a zone.
  source: https://developers.cloudflare.com/ai-crawl-control/
  tier: all
- id: pay-per-crawl
  name: Pay per crawl
  description: >-
    AI crawlers either present payment intent in request headers or get an HTTP 402 with a price per zone.
    Cloudflare is the merchant of record. Closed beta.
  source: https://developers.cloudflare.com/ai-crawl-control/features/pay-per-crawl/what-is-pay-per-crawl/
  tier: unknown
- id: x402-payments
  name: x402 paywalls for Workers and MCP tools
  description: >-
    x402 middleware (x402-hono) and an Agents SDK paidTool helper gate Worker routes and MCP tools behind
    an HTTP 402 stablecoin payment, settled through a facilitator.
  source: https://developers.cloudflare.com/agents/agentic-payments/x402/
  tier: open-source
- id: monetization-gateway
  name: Monetization Gateway
  description: >-
    Edge-enforced x402 charging for pages, datasets, APIs and MCP tools behind Cloudflare, with per-endpoint
    pricing rules. Waitlist since July 2026.
  source: https://blog.cloudflare.com/monetization-gateway/
  tier: unknown
- id: web-bot-auth
  name: Web Bot Auth verification
  description: >-
    Cloudflare verifies bots that sign requests with HTTP Message Signatures against a key directory the
    bot operator hosts at /.well-known/http-message-signatures-directory.
  source: https://developers.cloudflare.com/bots/reference/bot-verification/web-bot-auth/
  tier: all
- id: ai-index
  name: AI Index (auto-generated llms.txt, MCP server, search API)
  description: >-
    Cloudflare builds an AI-optimised index of the domain and exposes an MCP server, llms.txt/llms-full.txt,
    a search API and robots.txt discoverability directives. Announced as a private beta in September 2025.
  source: https://blog.cloudflare.com/an-ai-index-for-all-our-customers/
  tier: unknown
- id: workers-remote-mcp
  name: Remote MCP servers on Workers (Agents SDK)
  description: >-
    The provider writes and deploys a remote MCP server as a Worker, from Cloudflare templates. It serves
    on workers.dev by default and supports OAuth-protected or authless servers.
  source: https://developers.cloudflare.com/agents/guides/remote-mcp-server/
  tier: all
- id: agents-sdk
  name: Agents SDK
  description: >-
    A stateful agent runtime on Durable Objects with MCP, browser automation, sandbox, scheduling and
    payments tools.
  source: https://developers.cloudflare.com/agents/
  tier: open-source
- id: workers-oauth-provider
  name: workers-oauth-provider library
  description: >-
    An OAuth 2.1 provider framework for Workers covering RFC 8414 metadata, RFC 9728 protected-resource
    metadata, RFC 7591 dynamic client registration, PKCE, token revocation and Client ID Metadata Documents.
  source: https://github.com/cloudflare/workers-oauth-provider
  tier: open-source
- id: api-shield-schema-learning
  name: API Shield schema learning and export
  description: >-
    API Shield learns API operations from live traffic and exports them as an OpenAPI 3.0.0 JSON file,
    optionally with rate-limit recommendations.
  source: >-
    https://developers.cloudflare.com/api-shield/management-and-monitoring/endpoint-management/schema-learning/
  tier: unknown
- id: api-shield-schema-validation
  name: API Shield schema validation, endpoint discovery, mTLS
  description: >-
    Enforces an uploaded OpenAPI schema at the edge, keeps an API inventory, and handles mutual TLS with
    a Cloudflare-managed CA.
  source: https://developers.cloudflare.com/api-shield/
  tier: unknown
- id: rate-limiting-rules
  name: WAF rate limiting rules
  description: >-
    Counts requests per characteristic and blocks, challenges or logs over the threshold. The documentation
    mentions no standard RateLimit response headers for customer zones.
  source: https://developers.cloudflare.com/waf/rate-limiting-rules/
  tier: all
- id: pages-hosting
  name: Cloudflare Pages
  description: >-
    Git- or upload-deployed full-stack sites on Cloudflare's network. Cloudflare now points new projects
    to Workers.
  source: https://developers.cloudflare.com/pages/
  tier: all
- id: access
  name: Cloudflare Access
  description: Zero Trust gating of applications by identity and device-posture policy.
  source: https://developers.cloudflare.com/cloudflare-one/access-controls/policies/
  tier: unknown
maps:
- feature: content-signals-robots
  check: consent_identity
  layer: agent_readiness
  provider_must: >-
    Turn on managed robots.txt for the zone that serves its own domain. The ContentSignal pointer, which
    targets that robots.txt, must also be declared in apis.yml. The enrichment pipeline declares it when
    its probe finds the directive.
  note: >-
    The dimension is binary and reads the pointer, not the policy. Cloudflare's default values earn the
    same 3 points as a policy the provider wrote itself. About 99 catalog providers already carry a ContentSignal
    pointer, most of them pointing at robots.txt, which is the shape this feature produces.
  points: 3
  baseline_pass_rate: 0.006
  cohort_pct: 0.0
  control_pct: 0.6
  delta_pp: -0.6
- feature: security-txt
  check: well_known_published
  layer: composite
  provider_must: >-
    Fill in the security.txt fields (Contact and Expires are required) for the zone that serves its primary
    domain. The pipeline's /.well-known/security.txt probe then saves the file under well-known/.
  conditional: true
  condition: >-
    Only when the provider's primary domain is proxied through the Cloudflare zone where the file is configured.
    A security.txt on a marketing zone does not cover an API served elsewhere.
  catalog_pass_rate: 0.005
  facet: discoverability
  points: 6
  baseline_pass_rate: 0.018
  cohort_pct: 3.0
  control_pct: 1.8
  delta_pp: 1.2
- feature: security-txt
  check: security_disclosure
  layer: composite
  credit: 0.5
  provider_must: >-
    Set the Policy field to a real disclosure policy. The security-program probe writes the Security pointer
    from a security.txt Policy/Contact hit.
  note: >-
    Discounted because the generator produces only the pointer file. The disclosure program it points
    to is the provider's own work, and a Contact-only file is thin evidence.
  catalog_pass_rate: 0.115
  facet: operational_transparency
  points: 4
  baseline_pass_rate: 0.269
  cohort_pct: 39.4
  control_pct: 26.8
  delta_pp: 12.6
- feature: ai-index
  check: llms_txt_published
  layer: composite
  credit: 0.25
  provider_must: >-
    Get into the AI Index beta, enable it on the domain, and let the llms.txt be harvested or declared
    as an LLMsTxt pointer.
  note: >-
    The file is generated by crawling the provider's pages, not written by the provider, so under rule
    2 it grades as the rubric's derived class (0.25). The feature was announced as a private beta in September
    2025, and no fetched page confirms general availability.
  catalog_pass_rate: 0.351
  facet: discoverability
  points: 4
  baseline_pass_rate: 0.65
  cohort_pct: 75.8
  control_pct: 65.0
  delta_pp: 10.8
- feature: workers-remote-mcp
  check: mcp_server
  layer: agent_readiness
  grade: templated
  conditional: true
  condition: >-
    Only if the provider writes an MCP server over its own API, deploys it (ideally on its own domain,
    not workers.dev) and the probe verifies it. Cloudflare supplies the runtime and starter templates,
    not the tools.
  note: >-
    Graded templated because the realistic path is a Cloudflare starter template filled with the provider's
    tools. A server the provider builds by hand and the probe verifies would grade higher, but that credit
    belongs to the provider's work, not to Workers. An AI Index auto-generated MCP server is one template
    across customers and would grade platform (0.25).
  points: 12
  baseline_pass_rate: 0.22
  cohort_pct: 17.0
  control_pct: 22.0
  delta_pp: -5.0
- feature: workers-oauth-provider
  check: protected_resource_metadata
  layer: agent_readiness
  grade: verified
  conditional: true
  condition: >-
    Only if the provider puts the library in front of its API or MCP server on a host the probe reads.
    The library then serves /.well-known/oauth-protected-resource naming its authorization server.
  points: 5
  baseline_pass_rate: 0.133
  cohort_pct: 27.3
  control_pct: 13.2
  delta_pp: 14.1
- feature: workers-oauth-provider
  check: dynamic_client_registration
  layer: agent_readiness
  conditional: true
  condition: >-
    Only when the library is deployed as the provider's authorization server with RFC 7591 registration
    enabled, so that registration_endpoint appears in the served RFC 8414 document.
  points: 6
  baseline_pass_rate: 0.134
  cohort_pct: 21.2
  control_pct: 13.4
  delta_pp: 7.8
- feature: workers-oauth-provider
  check: delegated_identity
  layer: agent_readiness
  grade: served
  conditional: true
  condition: >-
    Only when the provider's served RFC 8414 document from the library lists authorization_code, which
    requires the provider to have a user identity to delegate.
  points: 6
  baseline_pass_rate: 0.209
  cohort_pct: 37.9
  control_pct: 20.8
  delta_pp: 17.1
- feature: workers-oauth-provider
  check: auth_clarity
  layer: agent_readiness
  grade: served
  conditional: true
  condition: >-
    Only when the provider actually uses the library's served RFC 8414 document, carrying issuer, as its
    API's authorization server. Wrapping a key-authenticated API does not change what the API accepts.
  points: 10
  baseline_pass_rate: 0.474
  cohort_pct: 55.2
  control_pct: 47.4
  delta_pp: 7.8
- feature: api-shield-schema-learning
  check: contract_present
  layer: composite
  conditional: true
  creates_artifact: true
  condition: >-
    Only if the provider exports the learned schema, reviews it, and publishes it as its own contract.
    The learned profile inside API Shield is not published anywhere the score reads.
  note: >-
    Learned from observed traffic, so it is real rather than invented. The file is OpenAPI 3.0.0 with
    paths, methods and parameters but no summaries, descriptions or error schemas, so the contract-content
    checks will mostly fail against it.
  catalog_pass_rate: 0.31
  facet: contract_quality
  points: 20
  baseline_pass_rate: 0.949
  saturated: true
  saturated_note: >-
    95% of providers with a contract, docs and a reference already earn this; the vendor cannot move it
    for most of its buyers.
  cohort_pct: 86.1
  control_pct: 94.9
  delta_pp: -8.8
earns_nothing:
- feature: pay-per-crawl
  check: agentic_commerce
  layer: agent_readiness
  why: >-
    agentic_commerce reads a parsing /.well-known/ucp.json or acp.json. Pay per crawl is an HTTP 402 price
    exchange with Cloudflare as merchant of record, has no discovery document, and is in closed beta.
- feature: monetization-gateway
  check: agentic_commerce
  layer: agent_readiness
  why: >-
    Edge-enforced x402 charging publishes no UCP/ACP discovery document on any page fetched, and it is
    waitlist-only.
- feature: web-bot-auth
  check: consent_identity
  layer: agent_readiness
  why: >-
    Verifying signed bots is site-side. The WebBotAuth/HTTPMessageSignatures pointers in the catalog point
    at a /.well-known/http-message-signatures-directory, which only a bot OPERATOR publishes. A provider
    that also runs its own signing agent earns that itself.
- feature: rate-limiting-rules
  check: rate_limit_signal
  layer: agent_readiness
  why: >-
    The dimension reads RateLimit headers documented in the provider's OpenAPI, or a rate_limits artifact.
    Edge rules enforce limits but, per the docs, emit no standard headers for customer zones. The RateLimit
    headers Cloudflare introduced are on its own API.
- feature: api-shield-schema-validation
  check: contract_present
  layer: composite
  why: Validation consumes an OpenAPI the provider uploads. It enforces a contract and publishes none.
- feature: pages-hosting
  check: apis_json_self_hosted
  layer: composite
  why: >-
    Hosting a file on the provider's domain is what any static host does. The credit belongs to the apis.yml/llms.txt/well-known
    file the provider writes, not to the host.
- feature: access
  check: auth_clarity
  layer: agent_readiness
  why: >-
    Access gates applications for workforce identity. The fetched docs show no OAuth discovery or protected-resource
    document it serves for an API's external consumers.
- feature: ai-crawl-control
  check: consent_identity
  layer: agent_readiness
  why: >-
    Blocking and analytics are enforcement. Only the managed robots.txt Content-Signal (mapped above)
    is something the score reads.
out_of_reach:
  checks:
  - openapi_3_1
  - info_complete
  - operations_summary_coverage
  - operations_description_coverage
  - error_responses_documented
  - examples_present
  - sdk_count_1
  - sdk_count_3
  - cli_present
  - plans_present
  - pricing_link
  - status_page_present
  - change_log_present
  - agent_card
  - agentic_commerce
  - idempotency
  - error_semantics
  note: >-
    An edge network cannot write a provider's contract, SDKs, plans or operational pages. Its payment
    features stop at HTTP 402 and never produce the UCP/ACP document agentic_commerce reads.
unscored_practice:
- feature: markdown-for-agents
  why: >-
    Accept text/markdown negotiation at the edge is deployed on Pro and above at no cost, and no 0.22.0
    check reads it. The content-signal response header it injects by default is also unread, because the
    ContentSignal pointer targets robots.txt, not a header.
- feature: x402-payments
  why: >-
    x402-gated Workers routes and MCP tools are real agent-payment practice. agentic_commerce reads only
    UCP/ACP discovery documents, so an x402 paywall scores nothing at 0.22.0.
surface:
  discoverability:
    reachable: 7.0
    total: 54
  contract_quality:
    reachable: 20.0
    total: 211
  operational_transparency:
    reachable: 2.0
    total: 38
  agent_readiness:
    reachable: 36.2
    total: 139
hard_rule: >-
  A model, not a score. Adopting this vendor changes a provider's Kin Score only when the provider publishes
  the resulting artifacts on its own surface; nothing here writes a score, and no sponsorship or partnership
  can.
method: searched
source:
- https://blog.cloudflare.com/an-ai-index-for-all-our-customers/
- https://blog.cloudflare.com/monetization-gateway/
- https://blog.cloudflare.com/security-txt/
- https://developers.cloudflare.com/agents/
- https://developers.cloudflare.com/agents/agentic-payments/x402/
- https://developers.cloudflare.com/agents/guides/remote-mcp-server/
- https://developers.cloudflare.com/ai-crawl-control/
- https://developers.cloudflare.com/ai-crawl-control/features/pay-per-crawl/what-is-pay-per-crawl/
- https://developers.cloudflare.com/api-shield/
- >-
  https://developers.cloudflare.com/api-shield/management-and-monitoring/endpoint-management/schema-learning/
- https://developers.cloudflare.com/bots/additional-configurations/managed-robots-txt/
- https://developers.cloudflare.com/bots/reference/bot-verification/web-bot-auth/
- https://developers.cloudflare.com/cloudflare-one/access-controls/policies/
- https://developers.cloudflare.com/fundamentals/reference/markdown-for-agents/
- https://developers.cloudflare.com/pages/
- https://developers.cloudflare.com/waf/rate-limiting-rules/
- https://github.com/cloudflare/workers-oauth-provider
measured:
  cohort:
    method: vendors-catalog.json detections (CNAME / header / URL shape / markup), never a name match
    detected: 121
    in_baseline: 33
  control:
    basis: providers earning contract_present + documentation_present + api_reference_present, minus the
      cohort
    n: 5183
  metric: >-
    cohort_pct / control_pct = mean share of the check's points earned (derived and platform credit weighted),
    x100
  measured_on: '2026-09-25'
  composite_mean:
    cohort: 48.8
    control: 48.0
  agent_readiness_mean:
    cohort: 34.9
    control: 32.5
  status: measured
  caveat: >-
    A cohort delta is association, not cause: customers choose a vendor for reasons that also move their
    score. Read it beside the capability map, never instead of it.
simulation:
  simulated_on: '2026-09-25'
  rubric: 0.23.0
  population: providers publishing a contract (contract_present earned), replayable exactly
  providers: 8977
  providers_unreplayable: 987
  providers_moved: 8962
  conditional_rows: excluded (they depend on what the API already does)
  composite_lift:
    median: 0.7
    p75: 0.9
    p90: 0.9
    max: 0.9
    mean_among_movers: 0.7
  agent_readiness_lift:
    median: 2.2
    p75: 2.2
    p90: 2.5
    max: 2.6
    mean_among_movers: 2.2
  facet_lift_median_among_movers:
    discoverability: 1.9
    operational_transparency: 5.3
  composite_band_moves:
    thin -> developing: 254
    developing -> strong: 93
    emerging -> thin: 75
    strong -> exemplar: 25
    minimal -> emerging: 1
  agent_readiness_band_moves:
    agent-aware -> agent-ready: 760
    agent-ready -> agent-native: 89
  method: >-
    each provider's own kin/checks file, the vendor's maps at their stated credit, the scorer's composite
    formula; from -> to, nothing written