Public API drift index
Public APIs that publish an OpenAPI document, checked read-only against that document and ranked by hard violations: the mismatches a client generated from the spec would break on. Each row links to every request sent and every finding, with a fix for each side.
| # | API | Spec format | Operations observed | Hard violations | Body findings | Status | Note |
|---|---|---|---|---|---|---|---|
| 1 | Discourse (meta.discourse.org) | OpenAPI 3 | 21 of 95 | 106 | 126 | Drift confirmed | Confirmed: category custom_fields holds plugin keys the spec forbids; topic tags are objects where the spec says string… |
| 2 | Forgejo (codeberg.org, self-served Swagger 2.0) | Swagger 2.0 | 38 of 506 | 23 | 35 | Drift confirmed | Same drift as Gitea (Forgejo fork), plus repo.parent null on feeds. |
| 3 | Gitea (gitea.com, self-served Swagger 2.0) | Swagger 2.0 | 40 of 537 | 15 | 30 | Drift confirmed | Confirmed: activity feed comment and user objects are null; User carries an undocumented username field. |
| 4 | APIs.guru directory API | OpenAPI 3 | 6 of 7 | 13 | 13 | Drift confirmed | Confirmed: GET /specs/{provider}/{api}.json returns a version entry (added, info, link, openapiVer, swaggerUrl, updated… |
| 5 | Wikimedia REST API (en.wikipedia.org) | OpenAPI 3 | 19 of 49 | 8 | 20 | Drift confirmed | Confirmed: media-list items carry undocumented leadImage and srcset; /page/lint returns an array where the spec says ob… |
| 6 | GitLab REST API (gitlab.com, unauthenticated) | OpenAPI 3 | 15 of 1864 | 3 | 3 | Drift confirmed | Confirmed: GET /projects/{id}/issues and /repository/contributors return arrays where the generated spec says object; n… |
| 7 | PokeAPI | OpenAPI 3 | 30 of 102 | 1 | 1 | Drift confirmed | Confirmed: evolution-chain baby_trigger_item is null; the spec does not allow null. |
| 8 | Swagger Petstore (reference demo) | OpenAPI 3 | 8 of 19 | 1 | 1 | Drift confirmed | Confirmed: GET /user/login returns a plain-text session string under application/json. |
| 9 | Open Library | OpenAPI 3 | 20 of 58 | 1 | 0 | Drift confirmed | Confirmed: /health returns text/plain, documented as application/json; the spec's own example for search.json spellchec… |
| 10 | Open-Meteo forecast API | OpenAPI 3 | 1 of 1 | 0 | 1 | Clean | Minor: current_units.interval is undocumented. |
| 11 | GitHub REST API (unauthenticated) | OpenAPI 3 | 40 of 1254 | 0 | 0 | Incomplete | First run: gist file language is null but not nullable; event payloads match two oneOf branches. |
| 12 | httpbin.org (self-served Swagger 2.0) | Swagger 2.0 | 20 of 78 | 0 | 0 | Clean | Confirmed: GET /cookies/set and /cookies/delete respond 302, documented as 200 only. Earlier input mistakes fixed. |
| 13 | Nager.Date public holidays | OpenAPI 3 | 8 of 8 | 0 | 0 | Clean | One 502 on LongWeekend during the run; transient, not drift. |
9 of 13 had at least one hard violation on a successful response, 171 distinct schema locations in total. Counts are distinct schema locations, not raw occurrences. Operations observed is the number of operations we got a response for, out of the operations in the spec.
Methodology
- Read-only. One
GETper operation, values from the spec’s own examples or a small hand-written set per API, a delay between requests, a named User-Agent, one polite retry on 429 and a hard request budget. Nothing is created, changed or deleted. - Unauthenticated. No credentials were sent, so the undocumented-status count in the per-API findings mostly reflects our own unauthenticated requests (401, 403, 429), not drift. It is kept out of this table.
- Hand-reviewed. Every hard violation was read by a person before it counted; the reviewer note says what we confirmed. One false positive in our own validator was found this way, fixed, and every API re-verified from the recorded responses.
- Status. Drift confirmed: at least one hard violation on a successful response. Clean: successful responses, none with a hard violation; read it with the coverage column. Incomplete: fewer than half the requests succeeded, usually rate limiting, so no verdict. Not verified: nothing sent or nothing succeeded.
- A sample, not a rate. These APIs were chosen because they publish a spec. The ranking says which of them drifted furthest in the operations we could reach, not what fraction of all APIs drift.
- Swagger 2.0 documents were converted to OpenAPI 3 before checking; Swagger 2.0 cannot express null, so some null findings on those rows are a format limitation as much as drift.
How to get your API on the index
Run the tool against your own public API and its published spec, thenopen an issue with the report attached and the spec URL. We rerun it read-only from our side, review the findings by hand, and add the row at the next monthly update. Clean rows are welcome; the index is more useful with them.
npx apidrifter run https://api.example.com/openapi.json --base-url https://api.example.com --format markdown > report.mdIf a finding on an existing row is wrong, open an issue and we will correct it; if it is right, a pull request on the spec is the fastest way to move the row down.
Data
- summary.json: this table as data.
- Per-API findings and observations: every finding with both fixes, every request with its status.
- The write-up: what we counted, what we did not, and the APIs worth reading about.
- Markdown twin of this page, for agents.
The index updates monthly. Text and data are Apache-2.0; cite as API Drifter (2026), Public API drift index, https://apidrifter.com/drift-index/.