aigw documentation

aigw is an OpenAI-compatible AI gateway. Point any OpenAI SDK at a single base URL, use a virtual key, and reach every model behind one contract - with per-key budgets, DLP redaction, and a full audit log. Run it EU-sovereign or self-host the stateless gateway in your own VPC.

Everything below is real today. If a sentence here doesn’t match what’s in the app, tell us at support@aigw.app.

Get started

New here? The Quickstart takes you from zero to your first completion in a few copy-paste lines.

The short version:

1. Sign in at console.aigw.app - magic link, no password.
2. Connect a provider and enable models. Add your own provider key, then turn on the models this workspace may use.
3. Mint a virtual key. Give it a monthly budget and, optionally, an allow-list of models.
4. Point your SDK at the gateway. Set base_url to https://gateway.aigw.app/v1 and use the virtual key as api_key.
5. Call any enabled model. Change the model string to switch providers - your code doesn’t change.

What’s in the box

• Virtual keys & budgets - issue a key per team, project, or agent; cap spend in US dollars; restrict which models it can use.
• Providers & models - every provider behind one endpoint, on your own provider keys, with retries and failover between providers of the same kind.
• Budgets, DLP & audit - the governance layer: spend caps, PII redaction, a per-call request log, and a tamper-evident audit log of every change.

The endpoints

aigw speaks the OpenAI API. Today it serves:

• POST /v1/chat/completions - chat, including streaming.
• POST /v1/messages - the Anthropic Messages API, which is what Claude Code and the Anthropic SDKs call.
• POST /v1/embeddings - embeddings.
• GET /v1/models - the models your key may use.

OpenAI- and Anthropic-shaped providers are translated for you, so the same request reaches either with no code change.