Providers and Backends

Backends are reusable provider accounts or compatible model endpoints. Add them once in the console, then use them across any number of model groups.

Supported Provider Types

Built-In Providers

Use catalog entries for providers such as OpenAI, Anthropic, OpenRouter, DeepSeek, Z.AI, Minimax, and Novita AI.

Coding Plans

Add plan-based coding accounts and monitor their quota, reset windows, and fallback behavior.

Custom Endpoints

Connect private or third-party services that behave like OpenAI-compatible or Anthropic-compatible endpoints.

Shared Capacity

Reuse one backend in multiple model groups while still tracking usage and health per backend.

What the Backend Screen Shows

• Provider and model identity.
• Whether the backend is credit-based or plan-based.
• Remaining balance, quota, or reset window when available.
• Live concurrency and rate-limit pressure.
• Which model groups currently use the backend.

Best Practices

• Use clear names such as anthropic-prod, openai-backup, or glm-coding-plan.
• Keep production, staging, and partner accounts separate.
• Check the “used in” column before editing or removing a backend.
• For coding plans, pair the plan backend with a reliable metered fallback.