> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.jambonz.org/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.jambonz.org/_mcp/server.

# Anthropic

Anthropic's Claude models are widely regarded as the strongest option for instruction following and tool use. Pair Claude with a fast STT and TTS for voice agents that need to follow complex multi-turn flows reliably.

## Get credentials

1. Sign in at [https://console.anthropic.com](https://console.anthropic.com).
2. Open **API Keys** in the sidebar (direct link: [https://console.anthropic.com/settings/keys](https://console.anthropic.com/settings/keys)).
3. Click **Create Key**, name it, copy the `sk-ant-...` string.

> **Tip**
>
> Anthropic shows the key once. Store it before navigating away — you can't retrieve it later, only revoke.

## Configure in jambonz

In the portal: **Account → LLM Services → + Add LLM Service → Anthropic**.

**`API Key`** `string` — required

The `sk-ant-...` secret from the Anthropic console.

---

**`Base URL`** `string`

Defaults to `https://api.anthropic.com`. Override only for proxies or enterprise deployments.

---

Click **Test** to verify.

## Use in an agent verb

```js
session.agent({
  llm: {
    vendor: 'anthropic',
    model: 'claude-haiku-4-5-20251001',
    llmOptions: {
      systemPrompt: 'You are a helpful voice assistant.',
    },
  },
  stt: { vendor: 'deepgram', language: 'en-US' },
  tts: { vendor: 'cartesia', voice: 'sonic-english' },
  turnDetection: 'krisp',
  bargeIn: { enable: true },
  actionHook: '/agent-complete',
}).send();
```

## Available Models

See Anthropic's [model catalog](https://docs.anthropic.com/en/docs/about-claude/models/overview) for the full list, capabilities, and current pricing. Common picks for voice agents:

| Model                       | When to use                                                    |
| --------------------------- | -------------------------------------------------------------- |
| `claude-haiku-4-5-20251001` | Default for voice — fast, cheap, tool-capable, vision          |
| `claude-sonnet-4-6`         | Mid-tier; better reasoning than Haiku, lower latency than Opus |
| `claude-opus-4-7`           | Highest-quality model; slower and more expensive               |

## Quirks & errors

> **Note**
>
> **Vision support varies by model.** Claude Sonnet and Opus accept image inputs; Claude Haiku 3.5 does not. Haiku 4.5 supports vision. If you need to pass images, double-check the model in the [Anthropic model comparison](https://docs.anthropic.com/en/docs/about-claude/models).

> **Warning**
>
> **`401 invalid_api_key`** — the key was revoked or has typos. Regenerate at [console.anthropic.com/settings/keys](https://console.anthropic.com/settings/keys).

> **Warning**
>
> **`429 rate_limit_error`** — you've hit Anthropic's per-minute or per-day rate limits. Reduce traffic or request a higher tier from Anthropic support. Note: Anthropic's rate limits are stricter than OpenAI's at equivalent tiers.