Skip to main content
Roo Code is an active fork of Cline. It is a multi-mode coding agent for VS Code, with modes such as Code, Architect, Ask, and Debug. Its OpenAI Compatible provider accepts a custom Base URL, so it can run on the ScitiX Model Inference API. Roo Code uses native tool calling only, with no XML prompt fallback. Choose a model that supports OpenAI-compatible function calling. glm-5.2 supports it. For the full list and pricing, see the model catalog.

​
Configure

Create an API Key on the API Keys page first.
  1. Click the Roo Code icon in the sidebar, then open Settings (⚙️) → Providers.
  2. Under API Provider, select OpenAI Compatible.
  3. Fill in the required fields.
  4. Keep Enable streaming on.
  5. Leave Enable R1 model parameters off. It is only for R1-style models such as QWQ.
  6. Optional: adjust Context Window and Max Output Tokens under the model settings to match your chosen model. If the context window exceeds the actual value for the selected model, context overflow may occur.
  7. Save the configuration.
Roo Code — OpenAI Compatible provider with ScitiX Base URL and model

​
Verify

Switch to Code mode and give Roo Code a task, such as “create test.txt in current folder and write in hello”. Roo Code plans the task and requests to write the file. Approve the action. A normal reply confirms that the connection works. Roo Code — creating a file from a coding task

​
FAQ

  • Stuck thinking, then errors / no tool calls: the model does not support native tool calling. Roo Code has no XML fallback, so switch to a model that supports it. glm-5.2 works.
  • 404: the Base URL ending in /v1 is enough. No extra path is needed.