One endpoint.
Many models.

VoidAI is a good router: one OpenAI-compatible endpoint in front of Claude Opus 5.5, Fable 5.1 and Sonnet 5.5. Point your client at https://api.voidai.dynv6.net/v1, keep your own key, and stream like normal. Workers at the edge, GCore in front, dynamic DNS behind a stable name.

See the models

Models, priced by the token.

A few of the routes on the shared key. Limits are monthly dollar allowances with 5-hour, weekly and monthly windows. This table is static for now — live numbers come later from the usage probe.

ModelViaIn / 1MOut / 1MMonthly allow.
claude-opus-5-5anthropic · messages$4.00$20.00$120
claude-fable-5-1anthropic · messages$10.00$50.00$60
claude-sonnet-5-5anthropic · messages$2.00$10.00$80
# chat completions, OpenAI-compatible curl https://api.voidai.dynv6.net/v1/chat/completions \ -H "Authorization: Bearer voidai-sk-YOU" \ -H "x-opencode-session: <stable-id-per-conversation>" \ -d '{"model":"claude-opus-5-5","stream":true,"messages":[…]}'

The route, in three hops.

Your client only ever sees the dynv6 name. Behind it, GCore absorbs the edge and the Worker does auth, per-user limits and the upstream call. Streaming responses pass through — the Worker never waits on tokens, so idle time costs nothing.

01 · Workers

Auth + proxy

Checks your VoidAI key, enforces your personal limit, forwards with the house key. →

02 · GCore

Edge + throttle

Static gets cached at the edge. /v1/* bypasses cache; heavy pulls slow to 64 KB/s after 512 KB. →

03 · dynv6

Stable name

One hostname for a moving setup. Point, forget, rotate origins without telling clients.