API reference
The base URL, request and response conventions, and where the generated OpenAPI document for every endpoint lives.
Everything the console does runs through one HTTP API at https://api.gigadrive.network. The generated OpenAPI document carries every path, parameter and schema. Dedicated sections in this reference cover authentication, scopes, error shapes and rate limits.
| What | Where |
|---|---|
| Base URL | https://api.gigadrive.network |
| Interactive reference | api.gigadrive.network/docs |
| OpenAPI document | api.gigadrive.network/docs/json |
| Credential | Authorization: Bearer <access token> on every request |
| Payloads | JSON in both directions |
The generated reference
/docs serves an interactive reference and /docs/json serves the OpenAPI 3.0.3 document behind it. Neither needs a credential. Both are built from the route definitions the service is running, so they describe the API that is deployed rather than a snapshot someone remembered to regenerate.
Go there for field names and response schemas. Come back here for the parts the document leaves out: how to get a token, which scope an endpoint wants, what an error body looks like, and how hard you may hit it.
Requests
Send Content-Type: application/json with a JSON body on POST, PUT and PATCH. The token endpoint is the one exception and takes application/x-www-form-urlencoded, as OAuth 2.0 requires.
Resource IDs in paths are UUIDs. Timestamps are ISO 8601 in UTC with milliseconds, such as 2026-06-03T12:00:00.000Z, in what you send as well as what you receive. An endpoint with nothing to return answers 204 No Content.
Pagination
Almost every list endpoint returns { items, total }. How you walk past the first page depends on which endpoint it is.
| Style | Query parameters | Extra response field | Endpoints |
|---|---|---|---|
| Keyset | from, limit (1 to 100, default 10) | none | GET /organizations, GET /applications, GET /deployments |
| Cursor | cursor, limit (up to 1000) | nextCursor | Storage objects, storage trash, request logs |
| Whole set | none | none | Everything else: environment variables, storage buckets, upload sessions, API keys, organization members, product access |
In the keyset style, send the id of the last item you received as from. In the cursor style, send back the nextCursor you received; it arrives as null on the final page.
On paginated endpoints total counts every record matching your filters and your token's reach, so a first page of 10 alongside total: 240 means 230 more are waiting. On whole-set endpoints it is the length of items.
Two endpoints sit outside all of this. Deployment build logs use offset and limit (default 100) and answer with totalItems, limit, offset and items, and accept createdAt[gt] for tailing. AI Gateway usage requests take a limit of up to 500 and return what fits.
Versioning
API paths carry no version segment: applications live at /applications, not /v1/applications. The AI Gateway is the exception and sits under /ai/v1, because it mirrors the OpenAI request and response layout its clients already expect.
The OpenAPI document declares its own info.version and is generated from the running service, so it moves whenever the API does. Gigadrive Network is in early access and the API is still moving, so pin your @gigadrive/sdk version and re-read the document after you raise it.