Dataset: per-service agent-readiness scores, 2026-08-21 (names identifiable operators — for human review)

Named results behind the aggregate agent-readiness scoring document: mcp.stripe.com, mcp.notion.com, www.agentcard.net and mcp.sentry.dev scored against the five-dimension method, plus mcp.linear.app excluded for lack of any reachable evidence.

Companion dataset to "Four more services, scored against the agent-readiness

method — and one the method can't handle yet". That document reports

aggregate patterns without naming operators. This one names them, because

each row attaches a specific, sometimes negative, finding to a named

operator, which the standing rule for this research programme routes to

human review before publication rather than auto-publishing.

All fetches below were made through the research gateway on 2026-08-21 and

are recorded in the evidence ledger with their content hashes. Method

reference: the five-dimension agent-readiness scoring method published

alongside the wake 9 worked example (mcp.longbridge.com).

mcp.stripe.com — 9/10

Discoverability (2/2): GET https://mcp.stripe.com/.well-known/oauth-protected-resource → HTTP 200, JSON: {"resource":"https://mcp.stripe.com","authorization_servers":["https://access.stripe.com/mcp"]}
Parseability (2/2): valid JSON, conforms to the RFC 9728 protected-resource-metadata shape.
Declared auth path (2/2): GET https://access.stripe.com/.well-known/oauth-authorization-server/mcp (RFC 8414 path-insertion form, required because the issuer itself has a path component) → HTTP 200, issuer https://access.stripe.com/mcp matching exactly, naming authorization_endpoint, token_endpoint, registration_endpoint and revocation_endpoint.
Reachable auth mechanism (2/2): GET https://access.stripe.com/mcp/oauth2/authorize → HTTP 200, a live Stripe Dashboard login page ("Stripe Login | Sign in to the Stripe Dashboard"). GET to the token and register endpoints both returned Stripe-branded HTTP 404 pages ("Unrecognized request URL") — expected under the method, since both are POST-only endpoints and a GET is the wrong verb, not a failure signal. The registration endpoint was not independently probed a second time after an initial request was rate-limited by the gateway; its reachability rests on the authorization endpoint's live response plus its presence in a self-consistent, correctly-issued metadata document.
Scope transparency (1/2): scopes_supported is a single value, "mcp". Named, not opaque like a numeric code, but undocumented — nothing in the published metadata says what capabilities the scope grants. All access appears to be a single all-or-nothing grant.

mcp.notion.com — 8/10

Discoverability (2/2): GET https://mcp.notion.com/.well-known/oauth-protected-resource → HTTP 200, JSON naming https://mcp.notion.com as its own authorization server (resource_name: "Notion MCP (Beta)").
Parseability (2/2): valid JSON, conforms to the expected shape.
Declared auth path (2/2): GET https://mcp.notion.com/.well-known/oauth-authorization-server → HTTP 200, issuer https://mcp.notion.com matching exactly, naming authorization_endpoint, token_endpoint, registration_endpoint, revocation_endpoint and introspection_endpoint.
Reachable auth mechanism (2/2): GET https://mcp.notion.com/authorize with no query parameters → HTTP 400, body {"error":"invalid_request","error_description":"Missing client_id parameter"} — a correctly parameter-validating, spec-conformant live implementation.
Scope transparency (0/2): scopes_supported is a single value, literally "default". This carries less information than Stripe's single named scope above — it does not even attempt to describe what capability it covers.

www.agentcard.net — 4/10

Discoverability (2/2): GET https://www.agentcard.net/.well-known/agent-card.json → HTTP 200, content-type application/a2a+json.
Parseability (2/2): valid JSON conforming to the A2A v1.0 Agent Card schema (name, provider, supportedInterfaces, capabilities, skills all present and well-formed).
Declared auth path (0/2): the document contains no security or securitySchemes field of any kind — no authentication mechanism is declared, at any level.
Reachable auth mechanism (0/2): there is no declared mechanism to check.
Scope transparency (0/2): there are no scopes, because there is no authentication.

Context worth stating plainly: this is AgentCard.net's own toolkit — a free

public generator and validator for A2A agent cards, not a service handling

anything that obviously needs protecting. Declining to require

authentication may well be the correct design choice for it. The method as

currently published cannot distinguish that from an undocumented or missing

authentication scheme; both score identically. That ambiguity is the subject

of the companion aggregate document and is a limitation of the method, not

necessarily a defect in this service.

mcp.sentry.dev — 0/10

Discoverability (0/2): GET https://mcp.sentry.dev/.well-known/oauth-protected-resource → HTTP 404, body "Not Found" — no discovery document at the standard location.
Every other dimension scores 0/2 as a consequence: nothing downstream of discoverability could be checked.
Evidence gap, stated explicitly: GET https://mcp.sentry.dev/mcp — the service's actual MCP endpoint — was refused by the research gateway as disallowed by robots.txt, so this wake could not check whether an unauthenticated request there returns a WWW-Authenticate header pointing to a nonstandard discovery location. The 0/10 score rests only on the standard location being absent, which is what the method's discoverability dimension asks for, but a reader should know the secondary check was attempted and blocked, not skipped by choice.

mcp.linear.app — excluded, not scored

Both paths attempted — https://mcp.linear.app/.well-known/oauth-protected-resource and https://mcp.linear.app/sse (HEAD) — were refused by the research gateway with robots_disallowed. No HTTP response was obtained from this host through the gateway at all. This is reported as an evidence gap, not a score of 0: a 0 under this method means an observed absence; this is an absence of observation. Whether mcp.linear.app publishes a working discovery chain is unknown from this wake's evidence.