Roo Code

Connect Roo Code with an OpenAI-compatible model that supports streamed native tool calls.

Install, configure, and test

  1. Open VS Code Extensions, search Roo Code, and install the official extension. Follow https://roocodeinc.github.io/Roo-Code/getting-started/installing/ if installing a VSIX instead. Open Roo Code's sidebar and click its settings gear. Create an API key in the routerra.ai console (https://routerra.ai/keys) and copy the exact ID of a model enabled for that key.

  2. Choose API Provider → OpenAI Compatible. Enter Base URL https://routerra.ai/v1, API Key YOUR_API_KEY, and Model ID YOUR_MODEL_ID in the model selector or custom ID field. This field takes the API base, not the full /v1/chat/completions URL. Do not select the dedicated OpenAI provider merely because the wire format is compatible.

  3. In Model Configuration, match Context Window and Max Output Tokens to the actual model. Enable Image Support only when supported; do not mark Computer Use available based only on text tool support. Input/output prices are Roo's local cost estimates, not gateway pricing controls. Save, and reload the VS Code window if the profile still uses stale settings.

  4. Start a new Code-mode task and ask Roo to read a small project file before attempting edits. Check native tool events and their results in the task timeline, then inspect the gateway log.

Current Roo documentation requires native OpenAI-compatible tool calling exclusively; there is no XML fallback. If chat text appears but tools fail, select a model and gateway route that support streamed native tool calls. Increasing context size will not add that capability. For 404 errors remove operation paths from Base URL; for model-not-found errors verify the selected profile's exact ID.