An AI gateway, operated like infrastructure
TokDash exists because buying model capacity should not require five vendor accounts, four currencies and a spreadsheet to work out who marked up what.
We aggregate capacity from 15 providers across the US, Europe and China and expose it through one OpenAI-compatible endpoint. Pricing is published per model, with the official list price and our price on the same row.
Requests are routed through Singapore and US-West, with automatic failover between regions. Endpoints are probed continuously; latency and uptime are published on the catalog.
We do not train models, do not sell data, and do not require Chinese residency or a local account to call Chinese models.
- 99.9%gateway uptime
- 42models, 4 modalities
- 2routing regions
Compliance & data handling
Prompt and completion content is not persisted. Operational metadata (latency, status, token counts) is retained for billing and incident review.
A data processing agreement is available for Volume and Enterprise customers.