> ## Documentation Index
> Fetch the complete documentation index at: https://docs.x402layer.cc/llms.txt
> Use this file to discover all available pages before exploring further.

# OpenAI-compatible Access

> Call your pod's agent from any OpenAI-compatible client. Create a key in Settings, point your SDK's base URL at the pod's /v1 endpoint, and every request runs one agent turn under the pod's wallet and spend policies.

Your pod exposes an **OpenAI-compatible endpoint**, so you can drive the agent from any OpenAI client, SDK, framework (LangChain, LlamaIndex, etc.), or router — no bespoke integration. Each request runs **one agent turn**, with the same model, wallet, autonomy, and spend policies as the dashboard Chat.

<Note>
  This is a **beta** feature and only appears once it's enabled for the platform. You'll find it in the pod workspace under **Settings → OpenAI-compatible access**.
</Note>

## 1. Create a key

In the pod's **Settings → OpenAI-compatible access**:

1. Copy the **base URL** — it looks like `https://compute.x402layer.cc/pods/{id}/v1`.
2. Give a key a name (e.g. "my app") and hit **Create key**. The key is shown **once** — copy it immediately.

You can create multiple keys and **revoke** any of them instantly if one leaks. Each key shows when it was last used and its per-day request count.

<Warning>
  Anyone with a key can instruct this agent within its current **autonomy and wallet settings**. Treat keys like secrets, and revoke immediately if one is exposed. See [Wallet](/cloud/pods/wallet) for spend caps.
</Warning>

## 2. Point any OpenAI client at it

Use the base URL as `baseURL` and your pod key as the API key.

```python theme={null}
from openai import OpenAI

client = OpenAI(
    base_url="https://compute.x402layer.cc/pods/{id}/v1",  # from Settings
    api_key="<your-pod-key>",                              # created above
)

resp = client.chat.completions.create(
    model="agent-pod",  # the pod runs its own configured model
    messages=[{"role": "user", "content": "What's on my schedule today?"}],
)
print(resp.choices[0].message.content)
```

```bash theme={null}
# Discover the model
curl https://compute.x402layer.cc/pods/{id}/v1/models \
  -H "Authorization: Bearer <your-pod-key>"

# Run one agent turn
curl https://compute.x402layer.cc/pods/{id}/v1/chat/completions \
  -H "Authorization: Bearer <your-pod-key>" \
  -H "Content-Type: application/json" \
  -d '{"model":"agent-pod","messages":[{"role":"user","content":"hi"}]}'
```

## What to expect

* **One turn per request** — a call runs the agent once (it may use its tools, wallet, skills, and memory), then returns the reply. It's the agent answering, not a raw model passthrough.
* **The pod's model** — the request uses whatever model the pod is set to (**Settings → Model** applies across chat, Telegram, and this API). See [Models](/cloud/pods/models).
* **Same policies as chat** — the pod's wallet, autonomy mode, and spend caps all apply to API-driven turns.

<Info>
  The pod's configured model is the source of truth — you don't select an arbitrary model per request the way you would against a raw provider. Set the model once in **Settings → Model**.
</Info>

## Full API spec

The examples above are the essentials. For the complete endpoint list, auth details, request/response schemas, and field-level reference, see the **[API reference](/api-reference/pods-adapter-chat)**.

## Related

<CardGroup cols={3}>
  <Card title="Models" icon="brain" href="/cloud/pods/models">Set the model this API uses.</Card>
  <Card title="Wallet" icon="wallet" href="/cloud/pods/wallet">Spend caps that apply to API turns.</Card>
  <Card title="API reference" icon="code" href="/api-reference/pods-adapter-chat">The full spec.</Card>
</CardGroup>