Inputs

Outcome: After this step, your buyer input form is defined and validated against your webhook contract.

The Inputs step defines what information buyers need to provide to run your AI Tool.

You'll create a dynamic form that collects exactly the data your automation needs to process.

Overview

Every AI Tool needs at least one input field. The inputs you define here become the form buyers fill out when purchasing your AI Tool.

The data they provide is sent to your webhook for processing.

Supported input types

Text fields

• text – Single-line text input (product names, keywords, etc.)
• textarea – Multi-line text input (descriptions, content, etc.)
• url – Website URL with validation
• tags – Comma-separated keywords or categories

Numbers and dates

• number – Numeric input with optional min/max limits
• date – Single date picker
• date_range – Start and end date selection

Choices

• select – Dropdown with predefined options
• multiselect – Multiple choice selection
• boolean – Yes/No or true/false checkbox

File uploads

• file_csv – CSV files for data analysis
• file_json – JSON files for structured data
• file_xlsx – Excel files for spreadsheet data
• file_pdf – PDF documents for analysis
• file_image – Images for visual analysis
• file_docx – Word documents for content extraction
• file_txt – Plain text files
• file_video – Video files (MP4, MOV, WebM, AVI) for video analysis or processing
• file_audio – Audio files (MP3, WAV, AAC, OGG) for audio analysis or processing

Field configuration

Key (required)

• Format: Must match ^[a-z][a-zA-Z0-9_]{1,39}$
• Rules: Start with lowercase letter, only letters/numbers/underscores, max 40 characters
• Examples: website_url, product_name, start_date
• Purpose: This becomes the property name in your webhook payload

Label and description

• Label – What buyers see as the field name
• Helper text – Additional guidance shown below the field
• Placeholder – Example text shown inside empty fields
• Required – Whether buyers must fill out this field

Validation options

• Number fields: Set min/max values and validation patterns
• Date fields: Configure date ranges and constraints
• Text fields: Pattern matching for specific formats
• Select fields: Define available options and default values

Multi-file options

For file input types, you can enable multiple file uploads:

• Multiple files – Allow buyers to upload more than one file
• Max files – Limit the number of files (e.g., 5 files maximum)
• Max bytes each – Size limit per file (e.g., 10MB per file)
• Accept formats – Automatically set based on file type selected

Mapping to webhook data

Each input key becomes a property in the JSON payload sent to your webhook. For example:

{
  "short_id": "abc123",
  "order_id": "order-uuid",
  "website_url": "https://example.com",
  "product_name": "Sample Product",
  "analysis_depth": 3,
  "include_images": true,
  "callback_url": "https://api.example.com/callback",
  "ts": 1736032405
}

Validation rules

• At least one input required – Every AI Tool needs input from buyers
• Unique keys – Each input key must be different
• Valid key format – Keys must follow the regex pattern above
• Select options required – Select and multiselect fields need predefined choices
• Number range validation – Min/max values must be logical
• Multi-file safeguards – File limits must be reasonable

Examples

SEO Analysis Agent

• website_url (url, required) – "Website to analyze"
• focus_keywords (tags, optional) – "Target keywords (comma-separated)"
• include_technical (boolean, default true) – "Include technical SEO analysis"

Product Research Agent

• search_term (text, required) – "Product or keyword to research"
• platform_context (text, optional) – e.g., "Amazon", "eBay", or "Walmart" (for output tailoring only; no login)
• date_range (date_range, optional) – "Historical data period"
• max_results (number, min: 10, max: 100) – "Maximum products to analyze"

Content Audit Agent

• content_files (file_pdf, multiple, max 5 files) – "Upload documents to audit"
• audit_type (multiselect, required) – Options: Grammar, SEO, Readability, Brand Voice
• target_audience (select, optional) – Options: General, Technical, Marketing

Authentication & Access Requirements

For AI Tools that need to access buyer accounts (like Amazon Ads or Amazon Seller Central), you can require buyers to connect their accounts in the "Authentication & Access Requirements" section of the Inputs step.

• Amazon Ads – Require buyers to connect their Amazon Ads account. Your automation receives a job_token that allows it to make API calls on the buyer's behalf.
• Amazon Seller Central (SP-API) – Require buyers to connect their Amazon Seller account for listings, catalog, orders aggregates, reports, and solicitations capabilities. (Financial-event data is not exposed today; see the Amazon Seller Central guide for the full scope.)
• Capabilities – All 42 Amazon Ads API capabilities (Sponsored Products, Sponsored Brands, Sponsored Display) and 11 Amazon SP-API capabilities are automatically enabled when you select the corresponding connection.
• Security – Buyers connect their own accounts securely through OAuth. Your automation only has access to the specific capabilities you use.

Learn more about Amazon Ads integration → · Learn more about Amazon Seller Central integration →

Post to Amazon (Optional): Apply-to-Amazon Button

If your AI Tool produces output that buyers can push directly to their Amazon account (e.g., updated listings, new ad campaigns, image patches), you can enable the Apply to Amazon button. Buyers see a one-click button on the order page that pushes your tool's output to their Amazon account via the broker.

In the Inputs step wizard, the "Post to Amazon (Optional)" section (collapsed by default) lets you declare which broker verbs your agent will call. It only appears when you've already selected Amazon Seller Central and/or Amazon Ads in the "Authentication & Access Requirements" section above it. Inside, verbs are grouped by provider and split into two subheadings:

• Write actions – Capabilities that mutate the buyer's Amazon account. Each checked write verb shows a per-verb toggle: Show "Apply to Amazon" button after tool runs (defaults to ON when you check the verb). Examples available in the broker catalog:
  • listings.update – Push listing patches (titles, bullets, descriptions, attributes)
  • listings.update_images – Push image patches
  • listings.a_plus_generate – Push A+ Content documents
  • ads.campaigns.create – Create Sponsored Products campaigns
  • ads.keywords.update, ads.negative_keywords.create – Update keywords and apply negative-keyword lists
  • ads.sb.campaigns.create, ads.sb.keywords.create, ads.sb.negative_keywords.create – Sponsored Brands v3 writes
  • ads.sd.campaigns.create, ads.sd.ad_groups.create, ads.sd.product_ads.create, ads.sd.targets.create, ads.sd.negative_targets.create – Sponsored Display writes
  • messaging.request_review – Trigger Amazon's Request-a-Review flow on a buyer order
• Read actions – Read-only capabilities listed in the verb catalog (e.g., listings.get, products.detail, catalog.search). Checking a read verb is declarative for catalog completeness; reads do not show an Apply button (no buyer account mutation involved).

Leave everything unchecked if your tool delivers content for the buyer to review manually. Existing tools default to no Apply button.

How it works:

1. You open the "Post to Amazon (Optional)" section and check the verbs your tool uses (read + write) in the wizard.
2. Admin reviews your tool. Tools that declare any write verb require an explicit "I approve this tool's declared write actions" checkbox before approval is allowed.
3. After approval, the order detail page renders the Apply to Amazon button to buyers for any write verb where you left the Show "Apply to Amazon" button after tool runs toggle ON, provided your tool's callback also returned a matching structured_output envelope.
4. When the buyer clicks Apply, the broker validates that the requested action is in the order's allowed scope, then executes the write verb against the buyer's connected Amazon account.

Callback payload requirement: The Apply button only renders if your tool's callback returns a structured_output envelope inside the per-output metadata field of a v2 callback (api_version: "2"). The shape:

{
  "metadata": {
    "structured_output": {
      "action": "listings.update",
      "provider": "amazon",
      "payload": { "sku": "ABC-123", "title": "New title here" },
      "confirm": {
        "summary": "Update title for SKU ABC-123",
        "diff_preview": {
          "before": { "title": "Old title" },
          "after":  { "title": "New title here" }
        }
      }
    }
  }
}

action must match a write verb you declared in the wizard. provider must be "amazon" or "amazon_ads". payload must match the broker's Zod schema for that verb (field names matter; camelCase will fail validation at Apply time). confirm.summary and confirm.diff_preview are optional but recommended.

See the full structured_output reference in the Callback guide for all fields, examples, and the v1 vs v2 callback caveat.

Important: If you edit which verbs your tool declares on an already-approved tool, the platform routes the edit back through the normal draft, submit, then admin review flow before the new scope is honored.

Tips for success

• Keep inputs minimal – Only ask for what you absolutely need
• Use clear labels – Avoid technical jargon buyers won't understand
• Provide examples – Use placeholder text to show expected formats
• Set smart defaults – Pre-select common options when possible
• Group related fields – Use logical ordering in your input list
• Test your keys – Make sure they match what your automation expects

What's next?

After defining inputs, you'll move to the Output step where you specify what results buyers will receive and how your automation will deliver them.