Quick start

The gateway speaks the OpenAI API. Three steps: get a key, put it in your environment, send a request. No SDK is required — if you already use one, it keeps working with two fields changed.

1

Create an API key

In the console, open Get started (or any workspace's API Keys page) and mint a key. It starts with aigw_ and is shown exactly once — the gateway stores only a hash. Every key carries its own budget caps and rate limits.

2

Put it in your environment

Never paste the key into code — the snippets read it from AIGW_API_KEY.

shell
export AIGW_API_KEY=aigw_…your key…
3

Make your first request

POST to /v1/chat/completions with a bearer token — that is the whole interface, and the cURL tab below is all of it. Already on an OpenAI SDK? Two fields change, base_url and api_key, and everything else stays.

Base URLhttps://aigw.173.212.206.55.nip.io/v1

Includes /v1. Drop it only if your tool appends the version itself.

AuthAuthorization: Bearer $AIGW_API_KEY
Modelopenai/gpt-4o-mini

Nothing to install — any HTTP client works, and the cURL tab is the whole contract. The SDK tabs are for pointing code you already have at the gateway.

curl https://aigw.173.212.206.55.nip.io/v1/chat/completions \
  -H "Authorization: Bearer $AIGW_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "openai/gpt-4o-mini",
  "messages": [
    {
      "role": "user",
      "content": "Hello!"
    }
  ]
}'

The gateway speaks the OpenAI API, so anything that can POST JSON works — an SDK, a client library, or a bare request. Budgets, rate limits, guardrails and logging apply to every call either way. Anything the provider refuses (a rate limit, a context-length error) is returned to your app unchanged, with the provider named. That URL is the gateway’s data plane; this console is a separate service and is not on your request path.

That request was budget-checked, rate-limited, guardrail-screened, priced and logged — visible in the console's Logs the moment it lands. Model ids are provider/model (e.g. openai/gpt-4o-mini); GET /v1/models lists what your key can reach.

Chat Completions

POST/v1/chat/completionsOpenAI-compatible chat
GET/v1/modelsList available models

API Keys

POST/admin/keysCreate API key
GET/admin/keysList all keys
DELETE/admin/keys/{key_hash}Disable a key
POST/admin/keys/{key_hash}/rotateRotate a key
GET/admin/keys/{key_hash}/analyticsPer-key usage stats

Guardrails & Safety

POST/admin/guardrails/testTest text against guardrails
GET/admin/safety-packsList safety packs
POST/admin/safety-packs/{id}/applyApply a pack

Monitoring

GET/admin/dashboardOverview metrics
GET/admin/dashboard/v2Enhanced metrics
GET/admin/auditAudit log
GET/admin/tracesRequest traces
GET/admin/breakersCircuit breaker status
GET/admin/complianceCompliance report
GET/healthzHealth check (no auth)
GET/metricsPrometheus metrics (no auth)

Routing & Simulation

POST/admin/whatifWhat-if cost simulation
GET/admin/catalogModel catalog
POST/admin/catalog/discoverDiscover models

Teams & Governance

POST/admin/teamsCreate team
GET/admin/teamsList teams
POST/admin/alertsCreate alert rule
GET/admin/alertsList alerts
POST/admin/webhooksCreate webhook
GET/admin/webhooksList webhooks

Authentication

Client requests (chat, embeddings): Authorization: Bearer <api_key>

Admin requests (keys, audit, config): X-AIGW-Admin: <admin_token>

SSO: OIDC and SAML supported. Session cookies set after auth flow.

Geo routing: X-Data-Region: eu to enforce region-based routing.

OAuth injection: X-User-Token: <oauth_token> for identity injection into agent requests.