PagerDuty

refuse — PagerDuty's own words on every operation in this family: "This API is in Early Access and may change at any time. Contact your PagerDuty account team to request access." This replica has not requested access and cannot — the gate is a conversation with a vendor, not a plan — so PagerDuty's refusal to an unenrolled account IS the true answer. Serving a Contextual Data Platform schema would additionally assert an integration posture (ServiceNow CMDB sync) that this engineering org has never had

refuse — PagerDuty's own words on every operation in this family: "This API is in Early Access and may change at any time. Contact your PagerDuty account team to request access." This replica has not requested access and cannot — the gate is a conversation with a vendor, not a plan — so PagerDuty's refusal to an unenrolled account IS the true answer. Serving a Contextual Data Platform schema would additionally assert an integration posture (ServiceNow CMDB sync) that this engineering org has never had

refuse — PagerDuty's own words on every operation in this family: "This API is in Early Access and may change at any time. Contact your PagerDuty account team to request access." This replica has not requested access and cannot — the gate is a conversation with a vendor, not a plan — so PagerDuty's refusal to an unenrolled account IS the true answer. Serving a Contextual Data Platform schema would additionally assert an integration posture (ServiceNow CMDB sync) that this engineering org has never had

refuse — PagerDuty's own words on every operation in this family: "This API is in Early Access and may change at any time. Contact your PagerDuty account team to request access." This replica has not requested access and cannot — the gate is a conversation with a vendor, not a plan — so PagerDuty's refusal to an unenrolled account IS the true answer. Serving a Contextual Data Platform schema would additionally assert an integration posture (ServiceNow CMDB sync) that this engineering org has never had

refuse — PagerDuty's own words on every operation in this family: "This API is in Early Access and may change at any time. Contact your PagerDuty account team to request access." This replica has not requested access and cannot — the gate is a conversation with a vendor, not a plan — so PagerDuty's refusal to an unenrolled account IS the true answer. Serving a Contextual Data Platform schema would additionally assert an integration posture (ServiceNow CMDB sync) that this engineering org has never had

refuse — PagerDuty's own words on every operation in this family: "This API is in Early Access and may change at any time. Contact your PagerDuty account team to request access." This replica has not requested access and cannot — the gate is a conversation with a vendor, not a plan — so PagerDuty's refusal to an unenrolled account IS the true answer. Serving a Contextual Data Platform schema would additionally assert an integration posture (ServiceNow CMDB sync) that this engineering org has never had

refuse — live authentication machinery, not engineering-org data: an inventory of the OAuth tokens and browser sessions currently issued against the account is a security surface whose truthful content changes by the minute and whose simulated content would be a set of credentials that authenticate nothing. The canon records no session and no token, and a replica that invented one would be publishing credential-shaped strings

refuse — live authentication machinery, not engineering-org data: an inventory of the OAuth tokens and browser sessions currently issued against the account is a security surface whose truthful content changes by the minute and whose simulated content would be a set of credentials that authenticate nothing. The canon records no session and no token, and a replica that invented one would be publishing credential-shaped strings. This operation is ALSO `deprecated: true` upstream with a documented replacement (the List OAuth Delegations endpoint) — a retirement candidate recorded in the review doc, deliberately NOT armed in the gateway registry here because PagerDuty has published no removal date

the v3 rendering of the same two ScheduleOverride rows as OverrideShift; rotation_id and custom_shift_id are omitted rather than invented, this canon has neither entity

the document supports exactly one feature here (`aiops`) and says an unentitled account gets a warning alongside the data. NO AIOps name appears in the 53-name captured /abilities response this replica serves, so `enabled: true` would contradict the list published one path over (invariant #5) — the entitlement is READ from our own ability list rather than chosen. The document leaves `enabled` unstated for the unentitled case and `false` is the only reading consistent with "not entitled to use"; the ambiguity is recorded in DECISIONS 2026-09-02 [pagerduty/enablements]. `updated_at` is omitted because nobody has ever changed a setting nobody can change

NULL, and the null is the fact rather than a gap: a global impact threshold is something an admin SETS, setting one is a write, and the document's own example for an account that has not set one is `{"global_threshold": null}`. Naming a priority would assert an impact rule this account never configured — and every incident here would then read as impacting business services it has none of

the ONE incident type PagerDuty ships to every account. Unlike /abilities — where the 2026-09-01 refusal found the document enumerated nothing — this list's contents are stated twice in the pinned document: the 200 on this operation says "The default incident type will automatically return on this list", and `IncidentType.parent` says a type with no parent "is created under top level (`incident_default`)". So the account has exactly one type until somebody writes another, and writing is out. The two NAMES are PagerDuty's own vocabulary for its own built-in row; the document's EXAMPLE id (P123456) and its 2023 timestamps are NOT copied — the id is minted by the same pdId every other object uses and the timestamps are the org's own creation instant. NOT the documented 402: the captured ability list carries `enable_custom_fields_on_incidents` and `premium_custom_fields`, so refusing would contradict the list published one path over

refuse — live authentication machinery, not engineering-org data: an inventory of the OAuth tokens and browser sessions currently issued against the account is a security surface whose truthful content changes by the minute and whose simulated content would be a set of credentials that authenticate nothing. The canon records no session and no token, and a replica that invented one would be publishing credential-shaped strings

refuse — network-security configuration behind an enrollment gate this replica is not enrolled in: every operation in this family requires the `X-EARLY-ACCESS: ip-allow-lists` header AND an account enrolled in the IP Allow Lists Early Access program, and its own description limits it to "Account Owners, Global Admins, and Account API Keys". An allow-list of CIDR ranges is also a claim about network perimeter that a public sandbox has no business simulating

refuse — network-security configuration behind an enrollment gate this replica is not enrolled in: every operation in this family requires the `X-EARLY-ACCESS: ip-allow-lists` header AND an account enrolled in the IP Allow Lists Early Access program, and its own description limits it to "Account Owners, Global Admins, and Account API Keys". An allow-list of CIDR ranges is also a claim about network perimeter that a public sandbox has no business simulating

refuse — account-administration machinery this replica does not model and should not simulate: seat licensing and plan allocation are a BILLING relationship, and this canon models an engineering organization rather than a customer of PagerDuty. There is no canonical row that could answer without inventing a plan for olympus-labs — the exact objection the 2026-09-01 /abilities refusal raised and the capture later answered for THAT row only. PagerDuty's own 4xx is the true answer here, not our coverage 404

refuse — account session-security policy (absolute and idle session TTLs), which is administration rather than engineering data — and the document states PagerDuty's own answer for an account that has not configured one: "If no configurations exist, a 404 Not Found error will be returned. A Session Configuration needs to be created before it can be retrieved." Creating one is a write, so the 404 IS this account's true answer rather than a coverage gap

TRUE AGGREGATIONS over the incident's own timeline, not statistics invented at render time: every duration is arithmetic over `Incident.openedOffset`/`acknowledgedOffset`/`resolvedOffset`, every count is over the `Page` rows the policy actually sent, and the business/off/sleep interruption buckets are computed from each page's sent instant on the RESPONDER's own clock (`Person.utcOffsetMin`) at PagerDuty's published boundaries. Six declared fields stay ABSENT because the canon states no fact behind them — `auto_resolved`, `major`, `snoozed_seconds`, `user_defined_effort_seconds` and the two `joined_user_*` — and a flat object of optional integers is exactly where a plausible number would never be caught

canon's `Deployment` rows in PagerDuty's own vocabulary — its `related_change_events` operation defines a change event as "service changes such as deploys, build completion, and configuration changes", which is the entity the git canon already records: a real commit, to a real environment, by a real person. `custom_details.commit` is the SAME sha every git host serves for that deploy, so a responder can follow one change across providers (invariant #5). `integration`, `routing_key`, `source`, `links` and `images` are omitted because this universe has no inbound integration — the reason the renderer already omits `Service.integrations`. `change_events` is one of the 53 names in the captured /abilities response, so the surface and the ability list agree

the deploy that landed nearest before the pager went off, on a repo the incident's own service ships from — labelled `most_recent`, which is the ONE of PagerDuty's three correlation reasons this canon can state as a fact. `related_service` would need a service-dependency graph the canon carries no edges for and `intelligent` names a model's output, so neither is emitted rather than being attached to the same row

served VERBATIM from a captured response (packages/conformance/specs/pagerduty/captures/abilities.json, sha-pinned in coverage/provider-pins.yaml under pagerduty.captures-file): 53 ability names captured 2026-09-02 from api.pagerduty.com using PagerDuty's PUBLIC developer-docs sample credentials, so the evidence is reproducible by anyone rather than taken on trust. That capture is the exact unblock the 2026-09-01 refusal named. It reaches PagerDuty's DEMO account, which is feature-rich — the operation says an ability depends on 'your pricing plan or account state', so no single list is correct and a captured one beats a chosen one

hello-13's `features.vendorCatalog`: twelve FICTIONAL integration partners, structure in canon's VENDOR_CATALOG, names and directory copy in themes/greek/vendors.ts, baked into the artifact's `vendor` table at compile. This REPLACES the 2026-09-02-morning `empty` ruling, which was itself a knowing deviation — the founder's afternoon ruling took the third option over both serving nothing and printing real companies' names. A pre-catalogue artifact (both frozen g7 pins) still answers 200 + the empty envelope, deliberately, not a generation gap

refuse — the ONE audit operation the document gates explicitly, in its own words: "Only admins, account owners, or global API tokens on PagerDuty accounts with the 'Audit Trail' feature can access this endpoint." The account-wide trail is both admin-only and plan-gated, and no ability in the captured /abilities response names Audit Trail. The RESOURCE-scoped audit rows carry no such sentence and are classified `generate` below — the split is the document's, not ours