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

# Anthropic

> Connect Anthropic to SentiVue Talk

Anthropic provides the **language model** that decides what your agent says. It does not provide voice or transcription, so pair it with a provider for each of those.

## What you need

| Field   | Required |
| ------- | -------- |
| API key | Yes      |

## Getting your API key

1. Sign in at [console.anthropic.com](https://console.anthropic.com/settings/keys).
2. Create a new key.
3. Copy it immediately — the Console shows the key once and only a hint afterwards.

Keys begin with `sk-ant-api03-`.

<Note>
  Keys belong to a **workspace**. Rate limits are set per workspace, so a key from a workspace with low limits will throttle even when your organization has headroom to spare. If you expect volume, put the key in a workspace sized for it.
</Note>

Make sure the account has credit. A key with no billing set up authenticates but fails on the first call.

## Adding the connection

1. Go to **Connections** and choose **Anthropic**.
2. Paste the API key.
3. Save.

SentiVue Talk validates the key with Anthropic as you save, so a bad or expired key is caught immediately rather than on a live call.

## What you can select afterwards

Select Anthropic as your LLM provider on [Model Configurations](/talk/configurations/inference-providers), then pick a model.

Latency is heard as a pause on a call, so start with the fastest model and move up only if the agent struggles with your prompt. Test the difference on a real call before deciding.

The dropdown lists the versions available. If the one you want is not there, type its name in — the field accepts any model your account can reach. See Anthropic's [model list](https://docs.anthropic.com/en/docs/about-claude/models).

## Completing the pipeline

Anthropic covers the language model only. You also need:

* A [voice provider](/talk/configurations/voice) to speak the responses
* A [transcriber](/talk/configurations/transcriber) to turn caller speech into text

Each is a separate connection.

## Troubleshooting

<AccordionGroup>
  <Accordion title="The key is rejected when I save">
    Confirm you copied the whole key including the `sk-ant-api03-` prefix, that it has not been revoked in the Console, and that the account has credit.
  </Accordion>

  <Accordion title="Calls fail with a rate-limit error">
    Rate limits apply per workspace, not per organization. Check the limits on the workspace the key belongs to, and move the key to a workspace with higher limits if needed.
  </Accordion>

  <Accordion title="The agent feels slow to reply">
    Switch to Haiku. Opus is noticeably slower, and in a voice conversation that delay is heard as the agent hesitating.
  </Accordion>
</AccordionGroup>