Plug in your own API provider
Add a custom endpoint in the Console and call it via {identity}/{model}. Requests are forwarded directly to your provider — free, no balance deduction, bypassing the platform's routing and billing.
Connect in 3 steps
Add a provider on the Console's "Bring Your Own Endpoint" page
Fill in Base URL, API Key, and the model list
Call with {identity}/{model}
model: abc123/gpt-4oRequests forward directly to your provider — free, no balance deduction
Core features
Why use a custom endpoint?
Completely free
Requests forward directly — no balance deduction, no plan calls consumed
Privacy & security
API keys are encrypted at rest; requests forward to your provider via the platform
Multi-endpoint failover
Add multiple providers; try in order with automatic switching
Standard-compatible
Supports OpenAI / Anthropic standard interfaces — zero-code integration
Use cases
Use vLLM / GPUSTack or other self-hosted inference services via a custom endpoint, then use them seamlessly in Cherry Studio, Cursor and other tools.
Add multiple API providers (OpenAI official, Azure, resellers, etc.); the platform auto-fails over in order — whichever is available gets used.
Use a custom endpoint for high-frequency models (where you already have a provider contract), and platform routing for other models — both together cut costs.
Configure one platform address in your tools; mix custom and platform models without switching API URLs.
FAQ
Are custom endpoints charged?
No. Requests forward directly to your provider — no balance deduction, no plan calls, no platform billing. You only pay your own provider.
Is my API Key safe?
API keys are encrypted in the database; the frontend only shows a masked version (e.g. ****xxxx). Keys are decrypted server-side at forwarding time and never exposed to the client.
Which interface formats are supported?
OpenAI standard (/v1/chat/completions) and Anthropic standard (/v1/messages). The system picks the right endpoint path automatically based on your request protocol.
What if multiple providers have the same model?
The system tries them in the order you set in the Console. When the first provider fails, it auto-switches to the next for failover.
Will a bare model name (no identity prefix) hit my custom endpoint?
Yes. When the platform routes don't have that model, the system falls back to your custom endpoint. But we recommend using the identity/model format to explicitly route to your endpoint.
Connect your first custom endpoint
Go to the Console's "Bring Your Own Endpoint" page to add a provider
Go configure