> 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/from-openai-direct.md).

# From OpenAI Direct

Two lines of your code change. There is a step before them that is not code.

***

## What setup is

1. **Register your OpenAI key with TOLVYN**, once, on the **Providers** page (`/providers`). TOLVYN calls OpenAI with *your* key, so until it has one every proxied request returns `409 provider_credentials_missing`. This is a dashboard step, not a code change, and it is the one people miss.
2. **Point the base URL at the proxy** — `https://api.openai.com/v1` → `https://proxy.tolvyn.io/v1/proxy/openai/v1`. **The trailing `/v1` matters on both sides**: the OpenAI SDKs append `/chat/completions` to whatever base you give them.
3. **Use a TOLVYN key in place of your provider key** — `sk-...` → `tlv_live_...`. Your OpenAI key stays with TOLVYN and never goes in your application again.

Steps 2 and 3 are the two lines. Step 1 is why the first request after a five-minute migration sometimes returns a 409 instead of a completion.

Your prompts, models, parameters, and response format are completely unchanged. TOLVYN speaks OpenAI's API natively.

***

## Before / after

### Python

**Before:**

```python
from openai import OpenAI

client = OpenAI(api_key="sk-...")
```

**After:**

```python
from openai import OpenAI

client = OpenAI(
    api_key="tlv_live_...",
    base_url="https://proxy.tolvyn.io/v1/proxy/openai/v1",
)
```

Add your real OpenAI key once in the TOLVYN dashboard, on the **Providers** page (`/providers`). Your app code never holds a provider key again.

***

### Node.js

**Before:**

```js
import OpenAI from "openai";

const client = new OpenAI({ apiKey: "sk-..." });
```

**After:**

```js
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "tlv_live_...",
  baseURL: "https://proxy.tolvyn.io/v1/proxy/openai/v1",
});
```

***

### curl

**Before:**

```bash
curl https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"Hello"}]}'
```

**After:**

```bash
curl https://proxy.tolvyn.io/v1/proxy/openai/v1/chat/completions \
  -H "Authorization: Bearer tlv_live_..." \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"Hello"}]}'
```

***

### Raw HTTP

**Before:**

```
POST /v1/chat/completions HTTP/1.1
Host: api.openai.com
Authorization: Bearer sk-...
Content-Type: application/json
```

**After:**

```
POST /v1/proxy/openai/v1/chat/completions HTTP/1.1
Host: proxy.tolvyn.io
Authorization: Bearer tlv_live_...
Content-Type: application/json
```

***

## What you get immediately

Once the two lines are changed, TOLVYN automatically:

* Logs every request with model, tokens, cost, and latency
* Calculates cost in real-time using live pricing data
* Stores a tamper-evident audit ledger of all spend
* Lets you set budgets and get alerts when limits are approached

No configuration required. The dashboard is live the moment the first request goes through.

***

## Fail-open guarantee

If TOLVYN ever has an issue, requests automatically pass through to OpenAI. Your application never goes down because of TOLVYN. This is by design — TOLVYN is in your request path for observability, not for availability.

***

## Optional: add attribution headers

To see cost broken down by team, feature, or user in the dashboard:

```
X-Tolvyn-Team: backend-team
X-Tolvyn-Service: summarization
X-Tolvyn-Feature: report-gen
X-Tolvyn-User: user-123
X-Tolvyn-End-Customer: acme-corp
```

None of these are required. Add them incrementally as you need the visibility.