Public API

Free, no key required, CORS enabled. Rate limited per client. Please cache responses using the ETag we return.

Quick start

Fetch the two newest published Codex events:

curl 'https://tokenresets.com/api/v1/events?provider=openai-codex&limit=2'

Open the live JSON response. Read endpoints require no API key and allow cross-origin GET,HEAD, and OPTIONS requests.

Representative response

This abbreviated example is generated from the newest event in the same dataset as the website. Cursor text is descriptive; clients must reuse the opaque value returned by the live endpoint.

{
  "data": [
    {
      "id": "715142ee-d20a-47cf-8d58-dc20c2db9d7d",
      "slug": "openai-codex-hard-reset-2026-10-02-eac569d8a1",
      "provider": {
        "slug": "openai-codex",
        "name": "OpenAI Codex"
      },
      "event_type": "hard_reset",
      "status": "published",
      "title": "OpenAI Codex: Usage limits reset",
      "announced_at": "2026-10-02T21:18:48Z",
      "scope": {
        "products": [],
        "plans": [],
        "windows": [],
        "scope_evidence": "unknown"
      },
      "confidence": {
        "label": "reported",
        "score": 0.5,
        "reasons": [
          "allowlisted public catalog classifies this item as regular",
          "catalog report does not state affected plans",
          "published as reported catalog evidence; excluded from interval stats",
          "auto-published at confidence reported (0.5)"
        ]
      },
      "links": {
        "html": "/events/openai-codex-hard-reset-2026-10-02-eac569d8a1",
        "self": "/api/v1/events/715142ee-d20a-47cf-8d58-dc20c2db9d7d"
      }
    }
  ],
  "pagination": {
    "next_cursor": "opaque value returned when more rows exist",
    "has_more": true,
    "limit": 1,
    "total": 81
  },
  "meta": {
    "generated_at": "2026-10-03T03:11:38Z",
    "schema_version": "1.0",
    "non_additive": false,
    "non_additive_reason": null
  }
}

Endpoints

Method	Path	Description
GET	/api/v1/health	Service status and operating mode.
GET	/api/v1/providers	Published providers with their latest event and coverage.
GET	/api/v1/providers/{slug}	Provider detail with statistics, 26-week heatmap, latest event, and latest reset.
GET	/api/v1/providers/{slug}/events	Events for one provider.
GET	/api/v1/events	All published events, newest first.
GET	/api/v1/events/{id-or-slug}	Event detail with evidence and revision history.
POST	/api/v1/reactions	Anonymous reaction. One per browser per event.
GET	/api/v1/reset-wishes	Current anonymous reset-wish counts by provider.
POST	/api/v1/reset-wishes	Add one repeatable reset wish to a provider total.
GET	/api/v1/feed.rss	RSS 2.0 feed.
GET	/api/v1/feed.atom	Atom 1.0 feed.

Rate limits and caching

Surface	Path	Fixed window	Notes
All public API routes	/api/v1/*	600 requests / 60 seconds	Base fixed-window limit per pseudonymous client key.
Event reactions	/api/v1/reactions	20 requests / 3600 seconds	Additional write limit; the base limit also applies.
Reset wishes	/api/v1/reset-wishes	120 requests / 3600 seconds	Additional POST limit; GET uses only the base limit.

Responses expose X-RateLimit-Limit, X-RateLimit-Remaining, and X-RateLimit-Reset. A rejected request returns 429, anapplication/problem+json body, and Retry-After.

Provider and event JSON is publicly cacheable for60 seconds. RSS and Atom feeds are cacheable for 30 seconds. Reset-wish reads and writes useCache-Control: no-store.

Query parameters

provider: Filter by provider slug.
event_type: Comma-separated event types.
from / to: RFC3339 bounds on announcement time. from is inclusive, to is exclusive.
limit: 1 to 100. Defaults to 50.
cursor: Opaque pagination cursor from the previous response.
minimum_confidence: Feeds only. Defaults to verified.

Response notes

Every response carries a meta object with the generation time and schema version.
Statistics include includedEventTypes, excludedCount, andexclusionReasons so you can see exactly what was counted.
Provider detail returns latest_event, latest_verified_reset, andlatest_published_reset. The last field may be a lower-confidence report.latest_reset remains a verified-only compatibility alias.
Fields the source did not state are null or empty withscope_evidence: "unknown". They are never inferred.
Pagination is keyset-based. Pass next_cursor back ascursor; do not construct cursors yourself.
Errors use application/problem+json and include arequest_id.

Conditional requests

curl -H 'If-None-Match: W/"abc123"' \
  https://tokenresets.com/api/v1/events

A matching ETag returns 304 Not Modified with no body. The ETag is computed from the content only, so it does not change just because you asked again.