HFAI Cloud · integrations
One endpoint.
Every client.
Eight ways in, all pointed at the same OpenAI-compatible base URL.
Eight clients,
one base URL.
A tool and a script for every stack,
each on its own page.
Every integration on this page is the same integration underneath: a client that already speaks the OpenAI chat-completions shape, given this base URL and a key minted in HFAI Cloud. None of the eight needs a plugin written for us, a second account or a separate balance. The tool's own configuration screen or file takes two values — the endpoint, https://api.heartflowai.net/v1, and the hfai_live_ key — and after that the model name in the request body decides which catalogue row answers. Each card opens the page for that client, which shows the exact fields, file or command its maintainers document. The one-line description under each name is the summary its own page opens with, not a claim about speed or quality; the model pages carry the dated facts about the models themselves. Start with curl if you want to see the request before wiring a client.
- curlOne command, one key, no client library.
- Pythonbase_url, api_key, and no other change.
- NodebaseURL, apiKey, one constructor.
- LangChainbase_url and api_key on ChatOpenAI.
- ClineThe provider form, filled in for this endpoint.
- ContinueOne YAML block, four keys that matter.
- OpenCodeA provider entry in opencode.json.
- Roo CodeThree provider fields, and the agent runs on your balance.
Each client page names the fields, the file and the command its own docs use. The endpoint, the key prefix and the model identifiers are identical across all eight.
Wire one.
Keep your stack.
The base URL and the key stay the same between clients, so an agent, a script and a shell command can share one balance.