One endpoint for every model your team runs.
Thtolen is a private, OpenAI-compatible API gateway. It routes chat, embedding and batch traffic to leading model providers with streaming, automatic failover and per-key usage controls.
curl https://api.thtolen.top/v1/chat/completions \ -H "Authorization: Bearer $THTOLEN_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "auto", "stream": true, "messages": [{"role": "user", "content": "Hello"}] }' # data: {"choices":[{"delta":{"content":"Hi"}}]} # data: [DONE]
Built for production AI workloads
A single gateway between your applications and upstream model providers, operated from a US-West edge close to provider regions.
OpenAI-compatible
Drop-in /v1 endpoints for chat, responses, embeddings and files. Keep your existing SDKs.
Streaming-first
Server-sent events over long-lived connections, with low time-to-first-token and sessions up to 60 minutes.
Routing & failover
Route by model, latency or cost. Requests retry on an alternate provider when an upstream degrades.
Batch & large payloads
Upload datasets and long-context documents up to 100 MB per request, or queue batch jobs for off-peak hours.
Usage & budgets
Per-key metering, monthly budgets and rate limits, with usage exported per workspace.
No prompt retention
Request bodies are not stored. Audit logs record metadata only: key, model, tokens and latency.
System status
Current state of gateway components.
- API · us-west Operational
- Streaming (SSE) Operational
- Batch & files Operational
- Console Operational
Access is by invitation only
Public registration is closed while we onboard teams in small batches. If your organization already uses Thtolen, ask your workspace administrator for an invitation. Learn more