One endpoint.
Built for flow.

Models tuned for the way you code, metered to the token, behind one OpenAI-compatible URL. Drop it into Cursor and stay in the zone.

Open your console How it works
https://atlasgw.com/v1
Set as the API base URL. Authenticate with your key. That's the whole integration.

Counted, and accounted for.

Every token in, every token out — metered per key, billed in credits, visible in real time. No surprises at the end of the month.

Your own catalog

Each account sees only its entitled models — including private aliases built for a single team, tuned to a single task.

Self-service keys

Mint, label, disable, and delete API keys yourself. Keys are hashed at rest and shown exactly once.

Live usage & balance

Tokens by key, model, and day. Current balance and burn rate on one screen, exportable as CSV statements.

Built-in memory

Models can carry a long-term brain scoped to your account — read-only reference or fully writable, your call.

Hardened by default.

Per-key rate limits, full audit trail, single-use login codes, and sessions that expire. The boring security work, already done.

Try it in the playground.

Sign in