---
title: Agent Mode: When Your Prompts Need Superpowers
description: Agent Mode combines web search, media generation, and multi-step workflows into one execution type.
date: January 16, 2026
author: Robert Soares
category: feature
---

You've got Standard mode. Deep Dive for research. Quick Report for faster research.

And then there's Agent Mode.

## What's Agent Mode?

Agent Mode gives your prompts access to multiple tools at once:
- Web search
- Media generation
- Workflow capabilities
- Multi-step task handling

It's for prompts that need to do more than just generate text.

## When to Use It

### Complex Multi-Step Tasks

"Research competitor X, create a comparison table, generate a presentation slide image, and draft an email to the team."

Standard mode would give you text about this. Agent Mode actually does the steps.

### Tasks Requiring External Data + Creation

"Find the latest statistics on remote work trends and create an infographic summarizing them."

Needs web search (current data) + image generation (infographic). Agent Mode combines both.

### Dynamic Workflows

"Analyze this company's website, identify their value proposition, and create three different ad copy variations targeting different audiences."

Multiple distinct steps, each building on the previous. Agent Mode territory.

## How to Use It

When creating or editing an action:

1. Click **Add Action** (or edit existing)
2. Set **Execution Mode** to **Agent Mode**
3. Save

Now when you run that action, it executes with full agent capabilities.

## Agent Mode vs. Other Modes

| Mode | What It Does |
|------|--------------|
| **Standard** | Generates text response based on the model's knowledge |
| **Deep Dive** | Searches 100+ sources, takes 7-10 min, full citations |
| **Quick Report** | 2-5 page report from 20+ sources, under 2 min |
| **Agent Mode** | Web search + media tools + multi-step execution |

## What It Can Access

- **Web Search** - Current information from the internet
- **Image Generation** - Create visuals as part of the task
- **Video Generation** - Create video content
- **Audio Generation** - Create voice or music
- **Your Knowledge Map** - Your uploaded documents
- **Workflow Logic** - Chain multiple operations

All generated media appears directly in your chat—images, videos, and audio render inline so you can review and iterate without switching contexts.

## When NOT to Use It

**Simple questions.** "What's the capital of France?" doesn't need Agent Mode.

**Pure research.** Deep Dive or Quick Report are better for comprehensive research reports.

**Creative writing.** Standard mode with a good model handles this fine.

**Speed-critical tasks.** Agent Mode can take longer since it's doing more.

## Example Use Cases

### Marketing
"Research our top 3 competitors' latest product launches, summarize key features, and create a comparison infographic."

### Sales
"Look up this prospect's company, find recent news about them, identify potential pain points, and draft a personalized outreach email."

### Content Creation
"Find trending topics in [industry], pick the best one for our audience, create an outline, and generate a header image."

### Analysis + Action
"Analyze the data in my uploaded spreadsheet, identify the top 3 insights, and create a summary slide with a chart visualization."

## Agent Mode in Automated Workflows

Here's where it gets powerful: **your saved actions can use Agent Mode**.

Set any action's execution mode to Agent Mode, add it to a workflow folder, and hit play. Your workflow now has full agentic capabilities—researching, generating media, and executing multi-step tasks automatically.

Example workflow:
1. **Research competitor** (Agent Mode) → searches web, creates summary
2. **Generate comparison chart** (Agent Mode) → creates infographic inline
3. **Draft outreach email** (Standard) → writes personalized email

Each step executes in sequence, with media appearing directly in chat as it's generated. One click, full automation.

## Tips

- **Be specific about steps.** "Do X, then Y, then Z" is clearer than "handle X, Y, and Z."
- **Mention what tools you want used.** "...and create an image showing..." makes it clear you want media generation.
- **Give context.** The more the agent knows about your goal, the better it executes.
- **Review outputs.** Multi-step processes can compound errors. Check each step.
- **Mix execution modes in workflows.** Not every action needs Agent Mode—use it where it adds value.

## Credits

Agent Mode uses more credits than Standard because it's doing more work. Factor that into your usage.

---

Agent Mode is for when you need your AI to actually *do* things, not just *write about* doing things.

Use it when the task is complex enough to warrant it. Stick with simpler modes when it's not.