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

# AI Calls

> Create custom AI-powered responses with your own prompts.

## What it does

AI Calls send your input to an AI model with a custom prompt you define. Unlike Hermes' built-in AI formatting, AI Calls give you complete control over the prompt and output.

## AI providers

| Provider      | Models                                 | API Key Required                |
| ------------- | -------------------------------------- | ------------------------------- |
| **Hermes**    | GPT-OSS-120B                           | No (included with subscription) |
| **OpenAI**    | GPT-5, GPT-5 Pro                       | Yes                             |
| **Anthropic** | Claude Sonnet 4.5, Opus 4.5, Haiku 4.5 | Yes                             |
| **xAI**       | Grok 4, Grok 4.1 Fast                  | Yes                             |

***

## Configuration

| Setting             | Default              | Description                   |
| ------------------- | -------------------- | ----------------------------- |
| **Provider**        | Hermes               | Which AI service to use       |
| **Model**           | Default for provider | Specific model                |
| **Prompt Template** | Required             | Your custom prompt            |
| **Output Handling** | Inject               | What to do with the response  |
| **Streaming**       | Off                  | Show response as it generates |

### Advanced settings

| Setting                  | Default       | Description                            |
| ------------------------ | ------------- | -------------------------------------- |
| **Temperature**          | 0.7           | Creativity (0 = focused, 2 = creative) |
| **Max Tokens**           | Model default | Response length limit                  |
| **System Prompt**        | None          | Additional AI instructions             |
| **Use Vision**           | Off           | Include screenshot for context         |
| **Conversation History** | Off           | Remember previous messages             |

***

## Output handling

| Option             | What happens               |
| ------------------ | -------------------------- |
| **Inject**         | Insert response at cursor  |
| **Clipboard**      | Copy response to clipboard |
| **Notification**   | Show as macOS notification |
| **Popup**          | Display in floating bubble |
| **Popup + Inject** | Show popup and insert text |

You can select multiple options.

***

## Template variables

Use these in your prompt template:

| Variable          | Description                |
| ----------------- | -------------------------- |
| `{transcription}` | What you said              |
| `{selected_text}` | Currently highlighted text |

***

## Example prompts

### Summarize

**Trigger**: "summarize this"
**Prompt**:

```
Summarize the following text in 3 bullet points:

{selected_text}
```

### Translate

**Trigger**: "translate to spanish"
**Prompt**:

```
Translate the following to Spanish:

{transcription}
```

### Explain code

**Trigger**: "explain this code"
**Prompt**:

```
Explain this code simply, as if to a junior developer:

{selected_text}
```

### Draft email

**Trigger**: "draft email about"
**Prompt**:

```
Write a professional email about the following topic:

{transcription}

Keep it concise and friendly.
```

### Answer question

**Trigger**: "ask hermes"
**Prompt**:

```
Answer this question helpfully and concisely:

{transcription}
```

### Fix grammar

**Trigger**: "fix this"
**Prompt**:

```
Fix any grammar, spelling, or punctuation errors in this text. Only return the corrected text, nothing else:

{selected_text}
```

***

## Creating an AI Call trigger

<Steps>
  <Step title="Open Triggers">
    Go to **Settings → Triggers**.
  </Step>

  <Step title="Create new trigger">
    Click **+** or **New Trigger**.
  </Step>

  <Step title="Set the phrase">
    Enter your trigger phrase (e.g., "summarize this").
  </Step>

  <Step title="Choose matching mode">
    **Exact** if using selected text, **Prefix** if using voice input.
  </Step>

  <Step title="Select AI Call">
    Choose **AI Call** as the action type.
  </Step>

  <Step title="Choose provider and model">
    Select your AI provider. Add API key if required.
  </Step>

  <Step title="Write your prompt">
    Enter your prompt template using variables.
  </Step>

  <Step title="Set output handling">
    Choose what to do with the response.
  </Step>

  <Step title="Save">
    Click **Save**.
  </Step>
</Steps>

***

## Using your own API keys

To use OpenAI, Anthropic, or xAI:

1. Go to **Settings → Integrations**
2. Enter your API key for the provider
3. Keys are stored securely and never shared

***

## Vision context

When **Use Vision** is enabled, Hermes includes a screenshot with your prompt. This lets the AI understand what's on your screen.

**Use cases**:

* "What's wrong with this code?" (AI sees the code)
* "Summarize this page" (AI sees the content)
* "Help me fill out this form" (AI sees the form fields)

***

## Tips

* **Be specific in prompts** — Clear instructions get better results
* **Use system prompts** for persistent instructions (e.g., "Always respond in bullet points")
* **Test with Hermes AI first** — It's included, no API key needed
* **Start with low temperature** (0.3–0.5) for factual tasks
* **Use higher temperature** (0.7–1.0) for creative tasks

***

## Related

<CardGroup cols={2}>
  <Card title="Text Formatting" icon="pen" href="/triggers/types/text-formatting">
    Built-in edit and reply actions.
  </Card>

  <Card title="Variables" icon="code" href="/triggers/variables">
    All available template variables.
  </Card>
</CardGroup>