Setup guides, indexed by tool

The promise of an OpenAI-compatible endpoint is one URL for everything. The reality is four different configuration carriers: an environment variable, a project JSON file, a UI field, or a per-node setting. This index keeps the tool-specific guides together so the next integration starts from a known shape.

How to pick a starting point

If the tool reads an environment variable, the change is reversible and applies to every project at once. If it stores credentials in a UI or a per-node field, later environment edits are silently ignored — that single difference explains most "the base URL is set but requests still go to the vendor" reports. Start with the guide for your client, then confirm the request appears in the gateway usage log.

Browse the other clusters

Error index

Base URL setup

Cost control

Migration

All 57 guides in this cluster

FAQ

Which tools support a custom endpoint?

Any client that exposes a base URL, api_base or provider URL field can be pointed at an OpenAI-compatible route. This index covers the ones we have configured and verified, tool by tool.

Why does a tool ignore the environment variable I set?

Several clients persist the endpoint in their own config at first run, and that stored value wins over later environment changes. The per-tool guide says where the stored value lives and how to clear it.

Does the model list have to be entered by hand?

Sometimes. A few clients never call /v1/models and keep their own list, so a valid endpoint shows an empty picker until the model IDs are added manually.

Check the current rate for your model before you commit a workload.