> For the complete documentation index, see [llms.txt](https://docs.tolvyn.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.tolvyn.io/migrate-to-tolvyn/index.md).

# Overview

TOLVYN sits in front of any AI proxy or direct OpenAI integration: the request and response shapes are unchanged, so your call sites stay as they are. What changes is the base URL, the key you send, and one piece of setup — registering your provider key with TOLVYN. Pick your migration path:

| From                                                      | Time       | Difficulty |
| --------------------------------------------------------- | ---------- | ---------- |
| [Helicone](/migrate-to-tolvyn/from-helicone.md)           | 15 minutes | Easy       |
| [Portkey](/migrate-to-tolvyn/from-portkey.md)             | 15 minutes | Easy       |
| [OpenAI direct](/migrate-to-tolvyn/from-openai-direct.md) | 5 minutes  | Trivial    |

All migrations preserve your existing API calls. No prompt changes. No model changes. TOLVYN is fail-open: if the proxy is unreachable and you configured a provider key, the SDK retries directly against the provider. Fail-open covers connection failures, timeouts and `503` — not `500`, `502` or `504` — and a fallen-back request is not metered.