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

# OpenAI

> Connect WednesdayAI to GPT and Codex models via API key or Codex subscription OAuth.

# OpenAI

WednesdayAI supports OpenAI's GPT and Codex model families. Authenticate with an API key, or with OAuth for Codex subscribers.

## API key

Create a key at [platform.openai.com](https://platform.openai.com) and onboard:

```bash theme={"dark"}
openclaw onboard --auth-choice openai-api-key
```

Or set the key on the gateway host:

```bash theme={"dark"}
echo 'OPENAI_API_KEY=sk-proj-...' >> ~/.openclaw/.env
```

Rotation variables (`OPENAI_API_KEYS`, `OPENAI_API_KEY_1`, `OPENAI_API_KEY_2`, and the single-override `OPENCLAW_LIVE_OPENAI_KEY`) are also honoured — see [Authentication](/admin/gateway/authentication).

## Codex subscription (OAuth)

OpenAI Codex OAuth is explicitly supported for external tools like WednesdayAI:

```bash theme={"dark"}
openclaw onboard --auth-choice openai-codex
# or
openclaw models auth login --provider openai-codex
```

This opens a browser for the OAuth flow and stores the token in the auth profile store.

## Selecting a model

```json5 theme={"dark"}
{
  agents: {
    defaults: {
      model: { primary: "openai/gpt-5.2" },
    },
  },
}
```

The Codex subscription provider uses the `openai-codex/` prefix, e.g. `openai-codex/gpt-5.3-codex`. Run `openclaw models list --all` to see the full catalog, or `openclaw models status` for what your auth can reach.

## Transport

The default transport is `auto` (WebSocket-first, with SSE fallback). Override per model:

```json5 theme={"dark"}
{
  agents: {
    defaults: {
      models: {
        "openai/gpt-5.2": {
          params: { transport: "sse" }, // "sse" | "websocket" | "auto"
        },
      },
    },
  },
}
```

The OpenAI Responses WebSocket warm-up defaults to on via `params.openaiWsWarmup`. Override it explicitly per model:

```json5 theme={"dark"}
{ openaiWsWarmup: true }  // enable (default)
{ openaiWsWarmup: false } // disable
```

## Responses API compaction

WednesdayAI auto-enables OpenAI Responses API server-side compaction to manage long-running context windows. When context usage reaches the threshold, the server compacts prior turns before sending the next request.

```json5 theme={"dark"}
{
  models: {
    providers: {
      openai: {
        responsesServerCompaction: true, // default: true
        responsesCompactThreshold: 0.7,  // default: 0.7 (70% context fill)
      },
    },
  },
}
```

Set `responsesServerCompaction: false` to disable, or adjust `responsesCompactThreshold` (0–1) to control when compaction triggers.

## Troubleshooting

<AccordionGroup>
  <Accordion title="401 Unauthorized">
    The API key is invalid or revoked. Generate a new one and re-run onboarding, or update `~/.openclaw/.env`.
  </Accordion>

  <Accordion title="429 Rate limited">
    You hit your tier limits. Add rotation keys via `OPENAI_API_KEYS` or upgrade your plan.
  </Accordion>

  <Accordion title="Codex OAuth session expired">
    Re-run `openclaw models auth login --provider openai-codex`.
  </Accordion>

  <Accordion title="Model not available on this key">
    Some models require specific access tiers. Check your access at [platform.openai.com/limits](https://platform.openai.com/limits).
  </Accordion>
</AccordionGroup>