OpenAPI spec
The Rental API and the Provider API as one OpenAPI 3.1 file: import it into Postman or Insomnia, generate a client, or hand it to an agent framework.
The whole HTTP API for renting GPUs and running your own nodes is described in one file:
https://paraloncloud.com/openapi.json
It is OpenAPI 3.1: every endpoint on these pages, with its request body, its answer field by field, and its errors. It is public and needs no key to download; calling the API still does.
What it covers
| Tag | Endpoints | Scope |
|---|---|---|
| Rentals | GET /gpus, GET /templates, POST /rentals, GET /rentals, GET /rentals/{id}, DELETE /rentals/{id} | rental (except /gpus and /templates) |
| SSH keys | GET /ssh-keys, POST /ssh-keys, DELETE /ssh-keys/{id} | rental |
| Account | GET /balance | any key |
| Provider | /nodes, /nodes/{id}, /nodes/{id}/command, /nodes/{id}/access, /nodes/{id}/reclaim, /earnings | nodes |
The inference API is not in it: it is OpenAI-compatible, so OpenAI's own spec and SDKs work against https://ai.paraloncloud.com/v1.
Use it
Postman or Insomnia. Import, then paste the URL above. Set the collection's auth to Bearer with your prlc_ key.
A typed client. Any OpenAPI generator works, for example:
npx @openapitools/openapi-generator-cli generate \
-i https://paraloncloud.com/openapi.json -g python -o paralon-client
Agent frameworks. Tools that turn an OpenAPI file into tool calls (LangChain's OpenAPI toolkit, UTCP clients, custom GPT actions) can read it directly. Give the agent a key with only the scopes it needs: a key with the rental scope spends credits. For Claude Code, Claude Desktop and Cursor the MCP server is the shorter path.
Keeping up
The file changes with the API, in the same release, and is checked against the live API after every deploy. info.version in the file says which version it is.