> ## 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.

# Anthropic

> Connect WednesdayAI to Claude — API key, OAuth token, or setup-token — plus thinking, prompt caching, and the 1M context window.

# Anthropic

WednesdayAI uses Anthropic's Claude models for its default AI. There are three ways to authenticate, and several Claude-specific options: thinking, prompt caching, and the 1M context window.

## Authentication

### Option A — API key (recommended)

Create a key in the Anthropic Console, then onboard:

```bash theme={"dark"}
openclaw onboard
# choose: Anthropic API key

# or non-interactive:
openclaw onboard --anthropic-api-key "$ANTHROPIC_API_KEY"
```

Or set the key on the gateway host so the daemon can read it:

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

### Option B — OAuth token env var

For scripted or containerised setups where you already hold an Anthropic OAuth token (`sk-ant-oat-...`):

```bash theme={"dark"}
export ANTHROPIC_OAUTH_TOKEN="sk-ant-oat-..."
```

`ANTHROPIC_OAUTH_TOKEN` is checked **before** `ANTHROPIC_API_KEY` — if both are set, the OAuth token wins. It bypasses the auth-profile store; `openclaw models status` shows it as **OAuth (env)**.

### Option C — Claude setup-token (subscription)

Setup-tokens are created by the Claude Code CLI on any machine:

```bash theme={"dark"}
claude setup-token
```

Paste it on the gateway host:

```bash theme={"dark"}
openclaw models auth paste-token --provider anthropic
```

<Warning>
  Anthropic has restricted subscription credential use outside Claude Code in the past. Setup-token support is technical compatibility only — verify Anthropic's current terms before relying on it. **API key auth is the safer, recommended path.**
</Warning>

## Selecting a model

```json5 theme={"dark"}
{
  agents: {
    defaults: {
      model: { primary: "anthropic/claude-opus-4-6" },
    },
  },
}
```

Common Claude model references:

| Model             | Reference                             |
| ----------------- | ------------------------------------- |
| Claude Sonnet 4.6 | `anthropic/claude-sonnet-4-6`         |
| Claude Opus 4.6   | `anthropic/claude-opus-4-6`           |
| Claude Opus 4.8   | `anthropic/claude-opus-4-8`           |
| Claude Haiku 4.5  | `anthropic/claude-haiku-4-5-20251001` |

## Thinking

Claude 4.6 models default to **adaptive** thinking when no explicit level is set. Override per message with `/think:<level>`, or pin it in model params:

```json5 theme={"dark"}
{
  agents: {
    defaults: {
      models: {
        "anthropic/claude-opus-4-6": {
          params: { thinking: "high" },
        },
      },
    },
  },
}
```

## Prompt caching

Prompt caching reuses the cache for repeated system-prompt prefixes, cutting cost and latency. It is **API-only** — subscription/OAuth auth does not honour cache settings. With an API key, WednesdayAI applies `cacheRetention: "short"` automatically; override it via model params:

```json5 theme={"dark"}
{
  agents: {
    defaults: {
      models: {
        "anthropic/claude-opus-4-6": {
          params: { cacheRetention: "long" },
        },
      },
    },
  },
}
```

| Value   | Cache TTL  | Notes                    |
| ------- | ---------- | ------------------------ |
| `none`  | No caching |                          |
| `short` | 5 minutes  | Default for API-key auth |
| `long`  | 1 hour     | Extended cache           |

You can set a model-level baseline and override per agent via `agents.list[].params` — for example one agent on `long` and a bursty alerts agent on `none` to avoid cache-write costs.

## 1M context window (beta)

Anthropic's 1M context window is beta-gated. Enable it per model with `params.context1m: true` for supported Opus/Sonnet models:

```json5 theme={"dark"}
{
  agents: {
    defaults: {
      models: {
        "anthropic/claude-opus-4-6": {
          params: { context1m: true },
        },
      },
    },
  },
}
```

<Note>
  The 1M context requires Anthropic to allow long-context usage on that credential (typically API-key billing). Anthropic rejects the `context-1m` beta on OAuth/subscription tokens, so WednesdayAI skips the beta header for OAuth auth automatically.
</Note>

## Environment variables

| Variable                                        | Description                                |
| ----------------------------------------------- | ------------------------------------------ |
| `ANTHROPIC_API_KEY`                             | Primary API key                            |
| `OPENCLAW_LIVE_ANTHROPIC_KEYS`                  | Comma-separated list of keys for rotation  |
| `ANTHROPIC_API_KEY_1`, `ANTHROPIC_API_KEY_2`, … | Individual numbered rotation keys          |
| `OPENCLAW_LIVE_ANTHROPIC_KEY`                   | Pin a specific key for the current session |
| `ANTHROPIC_OAUTH_TOKEN`                         | OAuth token (takes priority over API key)  |

## Troubleshooting

<AccordionGroup>
  <Accordion title="401 / token suddenly invalid">
    Subscription auth can expire or be revoked. Re-run `claude setup-token` and paste it on the gateway host with `openclaw models auth paste-token --provider anthropic`. For long-lived deployments, switch to an API key.
  </Accordion>

  <Accordion title="&#x22;This credential is only authorized for use with Claude Code&#x22;">
    Anthropic blocked the subscription credential for non-Claude-Code use. Use an API key instead.
  </Accordion>

  <Accordion title="&#x22;No API key found for provider anthropic&#x22;">
    Auth is per agent — new agents do not inherit the main agent's keys. Re-run onboarding for that agent, or paste a token/key on the gateway host, then verify with `openclaw models status`.
  </Accordion>

  <Accordion title="No available auth profile (all in cooldown)">
    Check `openclaw models status --json` for `auth.unusableProfiles`. Add another Anthropic profile or wait for the cooldown to clear.
  </Accordion>
</AccordionGroup>