# Public jobs API

Read-only JSON API for open OpenTrain jobs: list, search, read one job, the changes feed and facets. No key, CORS open, 600 requests a minute and no daily limit.

Source: https://www.opentrain.ai/docs/developers/job-feeds/api/

The public jobs API answers anonymous `GET` requests with JSON. It needs no account, key or token, allows requests from any website (CORS `*`) and has no daily limit. It is separate from the [authenticated Public API](https://www.opentrain.ai/docs/developers/api-reference/overview/) that employers and agents use to post and manage jobs.

- **Base URL:** `https://app.opentrain.ai/api/public/v1`
- **OpenAPI file:** [`https://app.opentrain.ai/api/public/v1/jobs/openapi.json`](https://app.opentrain.ai/api/public/v1/jobs/openapi.json). It is the contract of record; point code tools and agents at it.

For every open job in one request, use a [bulk file](https://www.opentrain.ai/docs/developers/job-feeds/feed-fields/) instead of paging through the API.

## Endpoints

| Method and path | What it returns |
| --- | --- |
| `GET /jobs` | The newest open jobs, newest first. |
| `GET /jobs/search?q=` | Open jobs ranked by relevance to `q`. |
| `GET /jobs/{id}` | One open job. |
| `GET /jobs/changes?since=` | Jobs published, updated or closed since a time. |
| `GET /jobs/facets` | Category and domain slugs with job counts. |
| `GET /jobs/openapi.json` | The OpenAPI file. |

### List and search

```bash
curl "https://app.opentrain.ai/api/public/v1/jobs?view=compact&order=newest&limit=25"
curl "https://app.opentrain.ai/api/public/v1/jobs/search?q=python&view=compact&limit=25"
```

| Parameter | Use |
| --- | --- |
| `q` | Search only. Free text, 1 to 200 characters, for example `pharmacy AI training`. |
| `view` | `compact`: jobs without descriptions or HTML. Without `view`, `GET /jobs` returns the full job with HTML. |
| `order` | `GET /jobs` only. `newest`. |
| `category` | A category slug from `GET /jobs/facets`. |
| `domain` | A subject-matter domain slug from `GET /jobs/facets`. |
| `language` | A language the work needs, such as `en`, `es` or `fil`. |
| `country` | Two-letter country of the person, such as `US`. Jobs with no country restriction are included. |
| `payType` | `PAY_PER_HOUR`, `FIXED_PRICE` or `PAY_PER_LABEL`. |
| `limit` | Jobs per page with `view=compact`, 1 to 25. Default 10. A larger number returns 25. |
| `cursor` | The `nextCursor` from the previous page of the same request. |

The answer has `generatedAt`, `jobs` and `nextCursor`. `nextCursor` is `null` on the last page. Search pages through the top 100 results.

### Read one job

```bash
curl "https://app.opentrain.ai/api/public/v1/jobs/{id}?view=agent"
```

The id is the `id` from a list or search result, or the part after the last `--` of a job page URL. `view=agent` returns the job with a plain-text description (`descriptionText`) and no HTML. A closed, private or unknown job returns `404`; a closed job's body is `{"id": "…", "status": "closed"}`.

Each job's `url` is its job page, where a person reads the job and applies. Link it exactly as returned. `applyMode` is `opentrain` when the person applies on OpenTrain, or `signup_first` when the person signs up on OpenTrain first and then finishes on another site (`external` in the agent view).

### Changes feed

```bash
curl "https://app.opentrain.ai/api/public/v1/jobs/changes?since=2026-11-01T00:00:00Z"
```

Each entry in `changes` has `jobId`, `eventType` (`PUBLISHED`, `UPDATED` or `CLOSED`) and `occurredAt`. The answer stops at the time in `until`. To stay in sync, pass `until` back as the next `since`. If `until` is before now, ask again at once from that time.

### Facets

`GET /jobs/facets` returns `totalJobs` and the `categories` and `domains` with a job `count` for each `value`. Use the values as `category` and `domain` filters.

## Tag your links

In the full view, each job's `url` and `applyUrl` carry `ch=public_api` and a `sub=` label. Send `?sub=` with your site's name (letters, digits, `.`, `_` and `-`, up to 40 characters) and the links carry it, so people who come from your site are counted as yours. Without it, `sub` is `api`.

## Limits and errors

- 600 requests a minute per IP address, with no daily limit. Searches with `q` (on `GET /jobs/search`, `GET /jobs` and the MCP search tool together) also count against 120 a minute per IP address.
- Answers are cached at the edge, so repeat reads are fast.
- Every answer carries `X-RateLimit-Limit`. Too many requests get `429` with `Retry-After` in seconds, plus `X-RateLimit-Remaining`, `X-RateLimit-Reset` and `RateLimit-Policy`. Wait that long, then carry on.
- A bad parameter value, such as an unknown `payType`, or a missing `since` on the changes feed gets `400`. An unknown job id gets `404`.

See [Fair use and limits](https://www.opentrain.ai/docs/developers/job-feeds/fair-use/).

## Related pages

- [Job feeds and API](https://www.opentrain.ai/docs/developers/job-feeds/overview/)
- [Feed field reference](https://www.opentrain.ai/docs/developers/job-feeds/feed-fields/)
- [Job search MCP server](https://www.opentrain.ai/docs/developers/job-feeds/mcp/)
- [Job feeds change log](https://www.opentrain.ai/docs/developers/job-feeds/changelog/)
