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

# Agentic solution overview

> An in-editor AI assistant that edits funnels by prompt, rebuilds screens from references, and generates full funnels from scratch.

The FunnelFox Agentic solution is a set of AI capabilities built
into the Editor. It understands your product, your niche, and
what converts in your category, then helps you build around it.

The agent composes screens, funnels, and variations live on the
canvas. Anyone on your team can ship an on-brand funnel without
builder training.

<Frame>
  <img src="https://mintcdn.com/funnelfox/I3KaNrZOm2PPb7Dr/assets/ai-agent.png?fit=max&auto=format&n=I3KaNrZOm2PPb7Dr&q=85&s=2402d77ea9c0055c0039dd36c48861fa" width="1920" height="989" data-path="assets/ai-agent.png" />
</Frame>

## How Agentic solution works

Instead of clicking through funnel screens and adjusting elements one by one, you describe the outcome you want in plain English, and the agent applies the edits directly on the canvas. The agent understands your product, your niche, and what actually converts in your category.

Every run is grounded in two sources of context:

* **Project context**: Your product profile (identity, category, brand, voice, and audience). Auto-captured from your App Store, Google Play, or website link and editable from the Context tab.
* **Global market context**: Structures, flows, and pricing patterns that convert across 30+ verticals in web2app.

This means the agent doesn't start from a blank template. It picks screen structures, copy tone, and pricing defaults that fit your product and your niche, so a fitness funnel doesn't look like an AI companion funnel.

## Key features

<ParamField path="AI Chat in Editor" />

Edit any element, screen, or the whole funnel by prompt, in seconds. No builder
skills required. Describe the change and the agent applies it.

Learn more about [AI Chat in Editor](/agentic-solution/overview#ai-chat-in-editor).

<ParamField path="Build from a reference" />

Drop a screenshot of a paywall, onboarding, or quiz screen. The agent
recreates it with your copy, colors, and components.

Learn more about [building from a reference](/agentic-solution/overview#build-from-a-reference).

<ParamField path="Generate funnels from scratch" />

Tell the agent about your app and it composes a complete, on-brand
funnel grounded in real market benchmarks for your niche.

Learn more about [generating funnels from scratch](/agentic-solution/overview#generate-funnels-from-scratch).

<ParamField path="Analytics agentic chat" />

<Callout icon="rocket" color="#e18535">Coming soon!</Callout>

Ask questions about your data in natural language. The agent queries
your metrics and builds the dashboard you need, right inside the chat.

<ParamField path="AI-driven A/B testing" />

<Callout icon="rocket" color="#e18535">Coming soon!</Callout>

The agent proposes which experiments to run, sets them up, and reads
the results, grounded in your data and category benchmarks.

<ParamField path="AI tips on your metrics" />

<Callout icon="rocket" color="#e18535">Coming soon!</Callout>

Proactive, in-context suggestions across analytics and experiments.
The agent surfaces what's working, what's slipping, and what to try
next.

## Quickstart

Before you start, set up your [project context](#project-context) so
the agent has something to ground its edits on. Context isn't
required, but results are noticeably better with it.

<Steps>
  <Step title="Open the AI panel">
    Open a funnel in the Editor and you will see the **AI Chat** on the left.
  </Step>

  <Step title="Give the agent context">
    Click any screen on the canvas to focus the agent on the whole screen, or any screen element for a more specific target.

    You can also click the paperclip icon on the bottom left to attach a
    reference image, or simply drag and drop, or even paste it
    with **Cmd/Ctrl + V**.
  </Step>

  <Step title="Describe the change">
    Type your prompt in plain English and press **Enter**. Use **Shift + Enter**
    for a new line.
  </Step>

  <Step title="Watch the canvas update">
    The agent makes changes. Everything happens instantly inside the Editor.
  </Step>

  <Step title="Iterate or undo">
    Send a follow-up prompt, or click **Undo** at the top to
    roll back that turn only.
  </Step>
</Steps>

## AI Chat in Editor

AI Chat in Editor is the baseline capability of the agent. It lives
in the **AI Chat** to the left from the canvas and lets anyone on your team
change any part of a funnel in plain language, without opening a
settings panel, picking a component, or editing a style.

You describe
what you want, and the agent applies the change. With AI Chat in Editor you can:

* **Edit elements**: Change a single element by prompt.
* **Edit screens**: Reshape an existing screen or create a new one.
* **Edit whole funnels**: Apply broad changes across every screen.
* **Localize in one prompt**: Translate a screen or the whole funnel
  into another language.

To use AI Chat in Editor:

<Steps>
  <Step title="Switch to the new Builder">
    Open the Editor and click **New Builder** in the top left to switch to the new version.

    To switch back, open the dropdown with your funnel name at the top left and click **Switch to Old Builder**.

    <Frame>
      <img src="https://mintcdn.com/funnelfox/cH_Wrn8HuoXXb6zy/assets/switch-builder.gif?s=acaf6cc617a9bea3c028e0bfb1a9b48a" width="1908" height="984" data-path="assets/switch-builder.gif" />
    </Frame>
  </Step>

  <Step title="Pick a scope">
    Open the Editor and use the canvas to select an element, screen, or nothing if you want to update the whole funnel.

    <Frame>
      <img src="https://mintcdn.com/funnelfox/I3KaNrZOm2PPb7Dr/assets/agent-add-context.gif?s=5a5f39eb6b24639b7ea67892986d4ca7" width="1908" height="984" data-path="assets/agent-add-context.gif" />
    </Frame>
  </Step>

  <Step title="Describe the change">
    Type your prompt in plain English and press **Enter**. Everything happens instantly right on canvas.

    <Frame>
      <img src="https://mintcdn.com/funnelfox/I3KaNrZOm2PPb7Dr/assets/agent-add-prompt.gif?s=ede965535c923161d1624de3cfe25413" width="1908" height="984" data-path="assets/agent-add-prompt.gif" />
    </Frame>
  </Step>

  <Step title="Review, refine, or undo">
    Read the agent's summary. Send a follow-up prompt to iterate, or
    click **Undo** at the top of the AI Chat to
    roll back that turn.
  </Step>
</Steps>

<Info>While the agent is working, the send button becomes a stop icon. If you click it to abort, the canvas reverts to its pre-prompt state.</Info>

<AccordionGroup>
  <Accordion title="Example prompts">
    * "Reduce the quiz options to 6."
    * "Change the CTA to 'Start my plan' and make it bolder."
    * "Create a new screen with a title, subtitle, and five single-choice questions about sleep habits."
    * "Turn this quiz screen into a paywall with a 3-plan picker and a testimonials block."
    * "Make the tone more friendly and less clinical across the whole funnel."
    * "Replace 'weight loss' with 'fitness journey' everywhere."
    * "Localize the whole funnel into Spanish."
  </Accordion>
</AccordionGroup>

### Agent context

You can feed the agent a few kinds of context alongside your prompt:

* **Screens & elements**: Click them with your mouse to select. The agent becomes focused on the whole screen or just a specific element.

* **Image attachments**: Click the paperclip icon at the bottom left to attach PNG, JPG, GIF, WebP, SVG, or BMP
  files. You can also drag and drop them, or paste with
  **Cmd/Ctrl + V**.

* **Conversation history**: Every prompt in a session is sent with
  the prior messages, so the agent can follow up on earlier
  requests. Say `do the same on the next screen` and it works as
  you'd expect.

### Agent & model settings

Model selection is located at the bottom of your AI Chat. From there, you can select:

* **Editor model**: The model used for the main screen/element editing work.
* **Other skills model**: The model used for supporting skills (planning, image analysis, etc.)
* **GPT version**: Only when GPT is selected in the Editor or other skills models.

<Frame>
  <img src="https://mintcdn.com/funnelfox/I3KaNrZOm2PPb7Dr/assets/agent-models.png?fit=max&auto=format&n=I3KaNrZOm2PPb7Dr&q=85&s=8f96458d90af3d7e2b633ef0aca30981" width="1920" height="989" data-path="assets/agent-models.png" />
</Frame>

Agent settings live under the **Editor menu** at the top left:

* **Images**: Choose how image assets are generated on screen.
  * **Off**: Skip image generation, the fastest one.
  * **Gemini**: Highest quality, but slower.
  * **Flash**: Faster, good for iteration.
* **Build**: Choose how the agent builds your funnel.
  * **Parallel**: Generates elements simultaneously and is
    faster.
  * **Sequential**: Generates them one after another, each
    informed by the previous, so it's more deterministic and more coherent
    for tightly linked content.
* **Reasoning**:
  * **Fast**: For single-element tweaks, copy changes,
    and simple localization.
  * **Thinking**: For complex generations
    where the agent plans its approach before acting.
* **Style**: Customize image style when building from scratch.
  * **Auto**
  * **Realism**
  * **Vector illustrations**
  * **3D**
  * **2D**

<Frame>
  <img src="https://mintcdn.com/funnelfox/I3KaNrZOm2PPb7Dr/assets/agent-settings.png?fit=max&auto=format&n=I3KaNrZOm2PPb7Dr&q=85&s=7700417efc070813524ce04b433772eb" width="1920" height="989" data-path="assets/agent-settings.png" />
</Frame>

## Build from a reference

Hand the agent any visual reference (an old design, an inspiration
shot, a competitor screen), and get a fully rebranded, editor-ready
screen in one step. The agent reads the reference, interprets its
structure, and recomposes it on your brand: your colors, your
components, your tone of voice.

The output is a real, native FunnelFox screen: fully
editable, not a locked artifact.

To build from a reference:

<Steps>
  <Step title="Attach your reference">
    Click the paperclip icon at the bottom left to attach PNG, JPG, GIF, WebP, SVG, or BMP
    files. You can also drag and drop them, or paste with
    **Cmd/Ctrl + V**.

    Works perfectly with paywalls, onboarding, quizzes,
    loaders, and feedback screens.
  </Step>

  <Step title="Add a clarifying prompt (optional)">
    A prompt helps when the reference alone
    is ambiguous.

    For example, `Rebuild this as
            the first screen of my quiz` or `Use this as inspiration for
            the pricing screen`.
  </Step>

  <Step title="Review and refine">
    Review your outcomes and refine by prompt or manually, when needed.
  </Step>
</Steps>

<Callout icon="rocket" color="#e18535">**Paste a URL** and **Figma integration** are coming soon!</Callout>

## Generate funnels from scratch

Before you start, describe your app or paste an App Store, Google Play, or website
link in the [Project context](/agentic-solution/overview#project-context).

The agent composes a complete, on-brand funnel from
scratch: welcome, quiz, processing, authorization, paywall, and
final screens. Structure, pricing, and onboarding flow are chosen
from market benchmarks for your niche, not guessed from a template.

Once the context is added, generate a funnel from scratch:

<Steps>
  <Step title="Start a new funnel">
    Go to the [Funnels](https://app.funnelfox.com/funnels) page and click **New funnel**.
  </Step>

  <Step title="Set up the agent">
    Click the **Editor menu** at the top left and select how the agent builds your funnel and visual style, if needed.
  </Step>

  <Step title="Provide a prompt">
    Trigger agent with something like `build me a funnel from scratch` and press **Enter**.

    Generation typically takes 2–4 minutes.
  </Step>

  <Step title="Review and refine">
    Review your outcomes and refine by prompt or manually, when needed.
  </Step>
</Steps>

<Note>
  Every funnel is unique. The same project regenerating gets
  something new. The agent pattern-matches against live funnels
  across 30+ verticals and applies what's converting in your
  category right now.
</Note>

## Project context

Project context is your product profile (brand, audience, and
category) that grounds every generation and edit. You'll find it on the **Context** page of your FunnelFox dashboard.

What gets captured:

* Identity and category (name, description, vertical)
* Branding and voice (color palette, tone, positioning)
* Audience, features, reviews
* Terms & privacy
* Social media

To set up project context:

1. Go to the **Context** page of your FunnelFox dashboard.

2. Paste an App Store, Google Play, or website link.

3. Click **Analyze**.

<Frame>
  <img src="https://mintcdn.com/funnelfox/I3KaNrZOm2PPb7Dr/assets/agent-project-context.gif?s=2b720f76fcb6c558c3ea7b1fe3d79cc8" width="1908" height="984" data-path="assets/agent-project-context.gif" />
</Frame>

The agent scrapes
the page and builds your project profile.

Click **Re-analyze** at any time to re-scrape your source. Useful
after a website refresh or store listing update.

<Callout icon="rocket" color="#e18535">
  Media extraction (images, screenshots, and videos
  straight from your product page) is coming soon!
</Callout>

## Usage limits

Each organization gets 1,000,000 tokens per month, shared across all
projects and all Agentic solution features. You can track consumption
in the [Usage tab](/dashboard/settings#usage) of your FunnelFox
dashboard Settings.