Cline

Cline is a VS Code extension, so its setup is a form rather than a config file. It reaches Inferray through the OpenAI Compatible provider.

Configure the provider

Open Cline's settings and fill in four fields:

Inferray answers GET /v1/models with the catalogue, so Cline's model list can populate itself. If it doesn't, type the id in — the models page has every one.

Set the model's limits

Under Model Configuration, Cline lets you declare what the model can do — max output tokens, context window, image support, and per-token price. Cline uses these to decide when to truncate a conversation and to estimate cost, so wrong numbers cost you either context or accuracy in its estimate. Each model's page has the real figures; the models list links to all of them.

The prices shown there are Inferray prices, which is what you actually pay, so entering those makes Cline's running cost estimate match your bill.

Pick a model that calls tools

Cline drives everything through tool calls, so a model without solid tool-calling support will stall part-way through a task. Every model in the catalogue supports tool calling — the capability is listed on each model page — but the stronger models hold up better over long agentic runs.

If a request fails

• 401 — the key is wrong, or it was pasted with whitespace around it
• 402 — the balance is too low; top up
• 404 naming the model — the id is misspelled, or that model isn't served on this endpoint. The response lists the ones that are.