1 · Get a key
Create a key in the dashboard at www.sideren.io/dashboard/keys. Keys start withsdn_ and are shown once at creation — store yours
in an environment variable, not in code.
2 · Set the base URL
Everything is served from one origin. OpenAI-style clients append/v1;
Anthropic-style clients use the bare origin (their SDKs add /v1/messages
themselves).
One origin, both dialects
3 · Make the first call
@sdn/gpt-oss-120b for any model in the catalog that your plan includes
(Plans & limits); in PAYG mode every model is available.
Closed-weight models are called by their plain name (gpt-6.1-sol); open-weight models
are called @sdn/<name>, as here. GET /v1/models lists these public ids, and the older
spellings (sdn-gpt-oss-120b, gpt-oss-120b) resolve to the same model. Set stream: true
for tokens as they are generated (see Streaming).
The Anthropic dialect works the same way: POST to
/v1/messages with your key in
x-api-key. Both dialects accept every model in the catalog. See the
API reference.Fast paths for coding agents
The setup helper configures popular coding agents in one command. Nothing to install, and safe to re-run:Claude Code
Four environment variables and Claude Code runs on the Sideren catalog.
Codex CLI
One provider block in config.toml, wire_api = responses.
Cursor
OpenAI-compatible base URL + key in the model settings.
All agents
OpenClaw, Hermes Agent, opencode, Cline and more — exact config for each.