Infrastructure
- Vercel
- Hosting, serverless functions, and web analytics. Requests pass through it in transit; TokenSpend does not store prompt or response content.
- Supabase
- Database. Account, workspace, key hash, billing ledger, and metadata-only request receipts.
- Stripe
- Payments. Billing details for credit top-ups; TokenSpend never sees full card numbers.
- Resend
- Transactional email. Email addresses and the content of sign-in and notice emails.
- PostHog
- Product analytics. Usage events from the website and dashboard.
Model providers
A request is sent to the provider that runs the model you chose, with your own key (BYOK) or with TokenSpend's managed access. These providers receive the prompt and return the response, under their own terms. Each model page lists the providers behind that model, every successful response names the one that served it in the x-tokenspend-provider-served header, and each receipt records it.
- Anthropic
- AWS Bedrock
- Azure OpenAI
- Baseten
- Cerebras
- DeepInfra
- Fireworks
- Google Cloud Vertex AI
- Groq
- Meta
- Novita
- NVIDIA
- OpenAI
- Together
- xAI
This list is generated from the providers TokenSpend can route to, so it changes when that does. Change history is kept in the legal version history.
Availability and support
During early access TokenSpend has no contractual uptime commitment or service credits. Live status and the validated check history are on the status page. If a provider stalls or fails, the router tries one more route; the reliability docs describe exactly when. An uptime commitment, support response times, and a data processing agreement can be agreed in an Enterprise order.
Enterprise orders, DPA and processor questions: sales@tokenspend.dev and privacy@tokenspend.dev.