Roo Code¶
Fill in three fields and every mode in Roo Code (Code, Ask, Architect, Debug and your own) runs on governed calls: checked against your policies, charged to your team and recorded in Request logs, on GPT, Claude or Gemini.
What it is¶
Roo Code is a coding agent for VS Code, which grew out of Cline. It works in steps, each one a model call with tools, and calls the model from your machine through the API Provider you choose.
The two settings¶
In Roo Code's settings, under API Provider:
| Field | Value |
|---|---|
| API Provider | OpenAI Compatible |
| Base URL | https://<your-gateway>/v1 |
| API Key | Your Vidai key (starts vk_) |
| Model | A model name such as gpt-4.1, claude-sonnet-4.5 or gemini-3.5-flash |
Under Model Configuration, set Context Window and Max Output Tokens to the model's own, and turn on
Image Support for models that read images. Roo Code sends the key as Authorization: Bearer. The gateway's address
is under Connect your app on the Keys page and in System › Settings › Gateway; the key comes from Keys › Issue
key (its secret is shown once).
A minimal call¶
- Fill in the fields above and save.
- Start a task with "Reply with exactly: ok".
- A short answer means the address and key are right.
Streaming¶
Roo Code streams every step; a stream from Claude or Gemini arrives as OpenAI chunks.
Tool calls¶
Roo Code relies on the model's native tool calling, with no text fallback, so pick a model that handles tools well. Tool calls travel in the OpenAI format and work on every provider the gateway translates to.
Reading what the gateway did¶
Roo Code doesn't show response headers. Look the calls up in Request logs: filter on Key, Model or
Path (/v1/chat/completions). One task shows as many calls, one per step.
Errors¶
Roo Code shows the gateway's error message on the step that failed. Common ones: 401 invalid_api_key, 403
model_not_allowed, 404 model_not_found (check Model), 403 content_guardrail_blocked, 429 (a limit on the
key or team). The full list is in Headers and errors.
Model names¶
Use the names your organisation approved under Models in the Control room.
What goes through the gateway¶
| Roo Code feature | Through the gateway |
|---|---|
| Every mode, built-in and custom, and every step of a task | Yes |
| Codebase indexing (embeddings) | Only if, under Codebase Indexing, Embedder Provider is OpenAI Compatible with Base URL https://<your-gateway>/v1, API Key your Vidai key, Model an approved OpenAI embedding model and Embedding Dimension set to its size |
Check it works in a minute¶
Start a task with "Reply with exactly: ok", then open Request logs a few seconds later and filter on your key.
Good to know¶
- Roo Code's repository was archived in May 2026: the extension keeps working as set up here, but gets no new versions. Cline takes the same settings.
- Autonomous tasks make many calls; a limit on the key keeps a runaway task's spend known.
- Issue one key per developer so cost reports show whose tasks made the calls.
- Roo Code's docs: OpenAI Compatible.