BYOK

Connect Your Own API Provider

Add your own endpoints in the console, call them with ID Code / Model Name (e.g., abc123/gpt-4o). Requests are forwarded directly to your provider - free, no balance deducted, bypassing platform routing and billing.

Three Steps to Connect

1

Add a provider in the console "BYOK" page

Fill in Base URL, API Key, and Model list

2

Use ID Code / Model Name to call

Model Name
model: abc123/gpt-4o
3

Requests go directly to your provider, free, no balance deducted

Bypasses platform routing, billing, model replacement, and compression

Core Features

Why use BYOK?

Completely Free

Requests forwarded directly, no balance deducted, no plan consumption

Privacy & Secure

API Key encrypted in storage, requests proxied through platform to your provider

Multi-Endpoint Failover

Add multiple providers, tried in order, automatic failover on failure

Standard API Compatible

Supports OpenAI / Anthropic standard APIs, zero-code integration

Use Cases

Self-hosted GPU Cluster

Use vLLM / GPUSTack or other self-hosted inference services, connect via BYOK, and use seamlessly in Cherry Studio, Cursor, and other tools.

Multi-Provider Aggregation

Add multiple API providers (OpenAI official, Azure, resellers, etc.), the platform automatically failovers in order, whichever works gets used.

Cost Optimization

Use BYOK for high-frequency models (where you already have provider contracts), and platform routing for supplementary models - dual approach to reduce costs.

Unified Entry Point

Configure just one platform address in your tools. Use your own models and platform models together without switching API addresses.

Frequently Asked Questions

Q

Is BYOK free?

Yes. Requests are forwarded directly to your provider - no balance deducted, no plan consumption, no platform billing. You only pay your own provider.

Q

Is my API Key secure?

API Keys are encrypted in the database. The frontend only shows a masked value (e.g. ****xxxx). Keys are decrypted server-side during request forwarding and never exposed to the client.

Q

Which API formats are supported?

Both OpenAI standard (/v1/chat/completions) and Anthropic standard (/v1/messages) are supported. The system automatically selects the correct endpoint path based on your request protocol.

Q

What if multiple providers have the same model?

The system tries them in the order set in your console settings. If the first provider fails, it automatically switches to the next, enabling failover.

Q

Will a bare model name (without ID Code prefix) use BYOK?

Yes. When the platform has no route for that model, the system automatically falls back to your BYOK providers. However, we recommend using the ID Code/Model Name format for explicit routing.

Connect Your First BYOK Endpoint

Go to the console BYOK page and add your provider configuration

Go to Settings