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

# Ai agent

# AI Agent

The **AI Agent** node calls a large language model to generate text or images. It supports any LLM provider, configurable system instructions, tool connections, and multi-modal operations.

<img src="https://mintcdn.com/subverse-611dde60/UZLalX8VX2i72yfQ/images/agentverse/ai-agent-ai-node.png?fit=max&auto=format&n=UZLalX8VX2i72yfQ&q=85&s=9cfb286c353162a5a29c830ff1abf41e" alt="AI Agent Node" width="806" height="1358" data-path="images/agentverse/ai-agent-ai-node.png" />

***

## Use Cases

* Generate summaries, classifications, or structured data from upstream inputs
* Answer questions based on dynamic context from other nodes
* Generate images from text prompts
* Use reasoning models for complex decision-making within an orchestration
* Build autonomous agents that use other nodes as tools

***

## Node Reference

### Parameters

| Parameter            | Type       | Required | Description                                                                            |
| -------------------- | ---------- | -------- | -------------------------------------------------------------------------------------- |
| **Provider**         | Select     | Yes      | The LLM provider to use: `OpenAI`, `Anthropic`, `Google`, or `Custom`                  |
| **Model**            | String     | Yes      | The model ID (e.g. `gpt-4o`, `claude-3-5-sonnet-20241022`, `gemini-1.5-pro`)           |
| **Operation**        | Select     | Yes      | `Generate a text response` or `Generate an image`                                      |
| **Credential**       | Credential | Yes      | API key credential for the selected provider                                           |
| **Instructions**     | String     | No       | System instructions that guide the AI behavior                                         |
| **Prompt**           | String     | Yes      | The user message or input to the model                                                 |
| **Reasoning Effort** | Select     | No       | `Low`, `Medium`, or `High` — applicable to GPT-5, o-series, and other reasoning models |
| **Service Tier**     | Select     | No       | The API service tier to use for the request                                            |
| **Max Tokens**       | Number     | No       | Maximum tokens in the response                                                         |
| **Temperature**      | Number     | No       | Sampling temperature (0–2). Higher = more creative.                                    |
| **Timeout**          | Number     | No       | Maximum time to wait for a response (in seconds)                                       |

***

## Providers

| Provider    | Credential Type        | Notes                                               |
| ----------- | ---------------------- | --------------------------------------------------- |
| `OpenAI`    | OpenAI API Key         | Supports GPT-4, GPT-4o, o1, DALL-E image generation |
| `Anthropic` | Anthropic API Key      | Supports Claude 3, Claude 4 models                  |
| `Google`    | Google AI API Key      | Supports Gemini 1.5, Gemini 2 models                |
| `Custom`    | Bearer Token / API Key | Use any OpenAI-compatible API endpoint              |

***

## Operations

### Generate a text response

Calls the model with system instructions and a user prompt. Returns the generated text response.

| Field                | Description                                                                          |
| -------------------- | ------------------------------------------------------------------------------------ |
| **Instructions**     | System instructions that set the model's persona or task context                     |
| **Prompt**           | The user message. Supports template expressions (e.g. `{{ webhook.body.message }}`). |
| **Reasoning Effort** | Controls the depth of reasoning for supported models (GPT-5, o-series).              |

### Generate Image

Generates an image from a text description (requires an image-capable model, e.g. DALL-E 3).

| Field          | Description                                             |
| -------------- | ------------------------------------------------------- |
| **Prompt**     | Text description of the image to generate               |
| **Image Size** | Output image dimensions (e.g. `1024x1024`, `1792x1024`) |
| **Quality**    | `Standard` or `HD`                                      |
| **Style**      | `Vivid` or `Natural`                                    |

***

## Tool Connections (AI Tool Edges)

Connect other nodes to the AI Agent via an **AI Tool** edge to give the agent the ability to call those nodes during reasoning. The agent autonomously decides when and how to use each tool.

<img src="https://mintlify.s3.us-west-1.amazonaws.com/subverse-611dde60/images/agentverse/ai-agent-tools.png" alt="AI Agent with Tools" />

Example: Connect an **Airtable** node and an **HTTP Request** node as tools. The agent will query Airtable or call the API as needed to answer the user's request.

***

## Output Data

For `Generate Text`:

```json theme={null}
{
  "text": "The primary benefits of AI automation include...",
  "model": "gpt-4o",
  "usage": {
    "promptTokens": 245,
    "completionTokens": 87,
    "totalTokens": 332
  },
  "finishReason": "stop"
}
```

Reference in downstream nodes:

```
{{ aiAgent.text }}
{{ aiAgent.usage.totalTokens }}
```

For `Generate Image`:

```json theme={null}
{
  "imageUrl": "https://oaidalleapiprodscus.blob.core.windows.net/...",
  "revisedPrompt": "..."
}
```

***

## Related Nodes

* [SubVerse Agents](./subverse-agents.md) — Run a full multi-turn SubVerse AI agent with channel support
* [Conditions](./conditions.md) — Branch based on AI output
* [Slack](./slack.md) — Send the generated text to a Slack channel