Integration guides
Cline
An open-source AI coding agent in VS Code that reads and writes across files and runs commands. It natively supports OpenAI-compatible providers — one base URL and key connects it to XylemNode.
- 01
Install the extension
In the editor, press Ctrl/Cmd+Shift+X to open the Extensions panel, search for Cline, and install the one by saoudrizwan. After installing, a Cline icon appears in the left activity bar — open it for the chat panel.
- 02
Choose the OpenAI-compatible provider
In the Cline panel, click the gear ⚙️ in the top-right to open settings, and switch the API Provider dropdown to 'OpenAI Compatible'. That's what lets you enter a custom address instead of going through official OpenAI.
- 03
Enter connection details
Fill in the three fields below. Note Cline's Base URL needs /v1 and should stop there — don't paste a full path like /chat/completions.
- Base URL
https://api.xylemnode.com/v1- API Key
sk-your-key- Model ID
gpt-5.6-sol
- 04
Verify and start
Click 'Verify' to confirm connectivity, then return to the chat panel to start. To change models later, just edit the Model ID — the Base URL and key stay untouched.
Two notes
1) If the model box doesn't auto-populate, just type the model name — go by the console's available list and mind the casing; a Model Not Found error usually means a wrong model name or group. 2) Model Configuration in settings tunes advanced items like max output and context window — leave them at defaults unless they don't match the model's actual capabilities.
Common questions
Cline reports Model Not Found, or the connection keeps failing
Cline's OpenAI Compatible provider has three required fields: Base URL, API Key and Model ID. The Model ID is typed in by hand — Cline does not fetch the available list for you — so it has to be the full identifier from the console, matching case, and then click Verify. The official troubleshooting list reads the same way: Model Not Found means that identifier does not exist behind the Base URL you entered, and a Connection Error means checking the address and whether your machine can reach it. One more thing: with a custom provider Cline knows nothing about the model's specs, so set the context window, max output tokens and the image / tool-use switches under Model Configuration. Left at their defaults, a long-context or image request can be rejected by Cline before it ever leaves your machine.