SHEET 07
Connection detail
OpenAI-compatiblecurl https://api.nerin.ai/v1/chat/completions \
-H "Authorization: Bearer $NERIN_API_KEY" \
-d '{
"model": "qwen3-coder-480b",
"messages": [{"role":"user","content":"Port this module to async."}]
}'
# the roles take the same key — hire the work, not the tokens
curl https://api.nerin.ai/v1/agents/code/runs \
-H "Authorization: Bearer $NERIN_API_KEY" \
-d '{
"repo": "git@github.com:acme/payments.git",
"task": "Fix the flaky timeout in the integration suite",
"budget_usd": 2.00
}'The inference API is a drop-in for the OpenAI schema — change the base URL and the model mark. The agent API is the part a model provider cannot give you: a run is a durable object with a budget, a transcript and artefacts you can inspect after the fact.
Base URLapi.nerin.ai/v1
AuthBearer key, scoped per project
CompatibilityDrop-in for the OpenAI chat schema
BillingPer unit, metered at the replica
Run budgetHard ceiling, enforced mid-run
DataNot trained on; not retained past the run