HFAI Cloud · integration
Roo Code
on HFAI Cloud.
Three provider fields, and the agent runs on your balance.
An open-source agent for VS Code.
Its own description,
its own provider page.
Roo Code introduces itself in its documentation as an open-source AI coding agent for VS Code, and it keeps a provider page of its own for OpenAI-compatible endpoints. That page names three settings and nothing else: the base URL for the endpoint, the API key that signs the request, and the model ID the agent should ask for. It also warns that the base URL is not the one a first-party account would use, which is exactly why the field exists. Checked against docs.roocode.com.
| Setting | What to put here |
|---|---|
| Base URL | https://api.heartflowai.net/v1 |
| API Key | hfai_live_… |
| Model ID | gpt-oss:120b |
Fill the three fields in the extension's provider settings and the agent starts drawing on the HFAI Cloud balance of the key you pasted. Nothing is patched, forked or rebuilt: the request is the ordinary chat completion, and the model ID is the only value that changes when you move on to another model.
Roo Code belongs to its own project and its own contributors. HeartFlowAI writes neither the extension nor its provider integrations, and the maintainers of Roo Code have not reviewed, approved or sponsored this page.
The same endpoint as every other client.
One key, one base URL,
one model string at a time.
The three fields above are the whole provider entry, so a second Roo Code profile can point at another model without touching the key or the URL.
The request shape and the error codes are in the API reference; the catalogue and its prices are on HFAI Cloud, and GPT-OSS 120B is the model the free tier covers.
One balance, several editors.
Roo Code is one of the editors this endpoint serves; the key and the model string carry over to the others unchanged.