CLI & IDE Compatibility Matrix
Reference for Openference client integration and verify flows.
* Codex
POST /v1/responses is adapted to chat completions for Chinese-model upstreams. See Chinese LLM Models.
Base URL convention
All clients should use:Usage counting
Requests count toward your plan’s window allowance when they succeed or fail with a client error (4xx). Upstream failures and capacity errors do not count.Cursor verify
Cursor’s Verify button callsGET {base_url}/models with the configured Bearer token. Openference returns an OpenAI-compatible model list filtered by token restrictions. Each model includes permission: [], root, and parent: null so Cursor’s client-side checks do not crash on missing fields.
Detailed guides
See the individual pages:- Cursor
- Chinese LLM Models
- OpenCode CLI
- Codex
- Kilo CLI
- Pi
- Open WebUI
- And others in the sidebar.
Model configuration tip
When a client lets you list specific models (e.g. OpenCode), only list the models you actually use. The model keys must exactly match the IDs returned byGET /v1/models.