> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bor-os.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Openai

# OpenAI

Use GPT models in BOR with an OpenAI API key. Adapter: `server/providers/openai.js`.

## Setup

1. Get an API key from the OpenAI platform (linked from onboarding).
2. In [onboarding](../getting-started/onboarding.md), pick **OpenAI**, paste your key, and choose a model.
3. BOR validates and you're ready.

## Choosing a model

Pick a **current** GPT model from OpenAI's live docs — model ids rotate, so avoid stale hardcoded ones. A vision-capable model is recommended so screenshot-based features work ([browser](../features/the-browser.md) previews, `locate_screen_element`, image attachments).

## Notes

* BOR's multimodal layer handles OpenAI image blocks and streaming via the shared provider utilities.
* Prompt-prefix caching is supported through a byte-stable prefix where the provider allows it.
* Edit `providers.openai` in `config.json` to change the key/model later, or ask BOR to switch.

## Troubleshooting

* **401** — bad key. **404 model** — model retired; pick a current one. **429** — rate-limited; slow down or upgrade your plan.
* See [Troubleshooting](../troubleshooting.md).