Skip to main content

​
Overview

The iPricing tool (part of the A-MINT service) transforms unstructured pricing pages into machine-readable Pricing2Yaml format. This enables H.A.R.V.E.Y. to perform rigorous analysis, validation, and optimization on any SaaS pricing model.

​
How iPricing Works

The extraction pipeline follows these steps:
1

URL Input

Provide a SaaS pricing page URL (e.g., https://buffer.com/pricing)
2

LLM-Powered Extraction

The A-MINT service uses OpenAI models to analyze the webpage content and extract:
  • Plan names and prices
  • Features and their types (DOMAIN, INTEGRATION, SUPPORT)
  • Usage limits (NUMERIC, BOOLEAN)
  • Plan relationships and constraints
3

Schema Generation

The extracted data is structured according to the Pricing2Yaml specification
4

Validation

The generated YAML is validated against the schema and can be checked for mathematical consistency
5

Caching

Results are cached to avoid re-extracting the same URL

​
Using iPricing via H.A.R.V.E.Y.

​
Automatic URL Detection

The easiest way to use iPricing is to include a URL directly in your question: 
What is the cheapest plan for Buffer (https://buffer.com/pricing) that includes 10 channels?
H.A.R.V.E.Y. automatically:
  1. Detects the URL in your question
  2. Calls the iPricing tool to extract the pricing model
  3. Adds the extracted YAML to the conversation context
  4. Proceeds to answer your question using the structured data

​
Manual URL Addition

You can also add URLs manually through the control panel:
1

Click 'Add URL'

In the context management section
2

Paste the pricing page URL

Enter the full URL including https://
3

Wait for transformation

The system displays a “pending” status while extracting. Once complete, it shows “done” with a ✓
4

Query the model

Ask questions about the pricing model

​
Tool Invocation (MCP)

If you’re using H.A.R.V.E.Y. as an MCP client or calling the MCP server directly, the iPricing tool accepts the following parameters:
​
pricing_url
string
required
The URL of the SaaS pricing page to extract
​
pricing_yaml
string
Optional: Provide existing YAML content instead of fetching from URL
​
refresh
boolean
default:"false"
Force re-extraction even if the URL is cached

​
Response Format

The tool returns a JSON object with: 
{
  "request": {
    "url": "https://buffer.com/pricing",
    "refresh": false
  },
  "pricing_yaml": "saasName: Buffer\nsyntaxVersion: '2.1'\n...",
  "source": "amint"
}
​
request
object
Echo of the input parameters
​
pricing_yaml
string
The extracted Pricing2Yaml document as a YAML string
​
source
string
Indicates whether the YAML came from amint (extraction) or upload (user-provided)

​
Pricing2Yaml Format

The extracted YAML follows the Pricing2Yaml specification. Here’s an example structure:
overleaf-extract.yaml
saasName: Overleaf - Individual
syntaxVersion: '2.1'
version: '2023-11-27'
createdAt: '2023-11-27'
currency: USD
variables: {}

features:
  githubIntegration:
    description: Link your Overleaf projects directly to a GitHub repository
    valueType: BOOLEAN
    defaultValue: false
    type: INTEGRATION
    integrationType: WEB_SAAS
  fullDocumentHistory:
    description: See all edits and who made every change
    valueType: BOOLEAN
    defaultValue: false
    type: INFORMATION
  prioritySupport:
    description: Priority and escalated support requests
    valueType: BOOLEAN
    defaultValue: false
    type: SUPPORT

usageLimits:
  maxCollaboratorsPerProject:
    description: Number of people you can invite per project
    valueType: NUMERIC
    defaultValue: 1
    unit: collaborator
    type: NON_RENEWABLE
  compileTimeoutLimit:
    description: Time allowed to compile your project
    valueType: NUMERIC
    defaultValue: 1
    unit: minute
    type: TIME_DRIVEN

plans:
  FREE:
    description: ''
    monthlyPrice: 0
    annualPrice: 0
    unit: /month
    features: null
    usageLimits: null
    price: 0
  
  STANDARD:
    description: ''
    monthlyPrice: 21
    annualPrice: 16.59
    unit: /month
    features:
      githubIntegration:
        value: true
      fullDocumentHistory:
        value: true
      prioritySupport:
        value: true
    usageLimits:
      maxCollaboratorsPerProject:
        value: 11
      compileTimeoutLimit:
        value: 4
    price: 21

​
Schema Components

  • saasName: Name of the SaaS product
  • syntaxVersion: Pricing2Yaml spec version (currently 2.1)
  • version: Version identifier for this pricing model
  • createdAt: Extraction timestamp
  • currency: Price currency (USD, EUR, etc.)
Each feature has:
  • description: Human-readable explanation
  • valueType: BOOLEAN, NUMERIC, TEXT
  • defaultValue: Base value (usually for free plan)
  • type: DOMAIN, INTEGRATION, SUPPORT, INFORMATION, MANAGEMENT, GUARANTEE, AUTOMATION, PAYMENT
  • integrationType (optional): API, WEB_SAAS, IDENTITY_PROVIDER, EXTENSION
Quantifiable constraints:
  • valueType: NUMERIC or BOOLEAN
  • defaultValue: Base limit
  • unit: Measurement unit (users, requests, GB, etc.)
  • type: RENEWABLE, NON_RENEWABLE, TIME_DRIVEN, RESPONSE_DRIVEN
  • linkedFeatures: Features that depend on this limit
Each plan specifies:
  • monthlyPrice: Price per month
  • annualPrice: Price per month if billed annually
  • unit: Pricing unit (usually /month)
  • features: Override feature values for this plan
  • usageLimits: Override limits for this plan
  • price: Effective price (usually equals monthlyPrice)

​
Extraction Best Practices

Use Official Pricing Pages

Extract from the official SaaS vendor pricing page, not third-party comparison sites.

Check for Completeness

After extraction, review the YAML to ensure all plans and features were captured correctly.

Refresh Stale Data

Use refresh: true to force re-extraction if pricing has changed since the last fetch.

Validate After Extraction

Use the validate tool to check the extracted model for mathematical consistency.

​
Caching Behavior

The MCP server caches extracted pricing models by URL:
  • Default TTL: Configured via CACHE_BACKEND and cache settings
  • Cache Key: pricing:<url>
  • Bypass Cache: Set refresh: true in the tool call
Caching reduces latency and API costs by avoiding redundant extractions. The cache is shared across all H.A.R.V.E.Y. sessions.

​
Configuration

The A-MINT service requires an OpenAI API key:
docker-compose.yml
a-mint-api:
  environment:
    - OPENAI_API_KEY=${AMINT_API_KEY}
    - OPENAI_API_KEYS=${AMINT_API_KEYS}
Set the key in your .env file:
.env
AMINT_API_KEY=sk-...
Without a valid OpenAI API key, the iPricing tool will fail. Ensure your key has sufficient quota and permissions.

​
Troubleshooting Extraction

​
Incomplete Feature Extraction

Problem: Not all features or plans are extracted Solutions:
  • Verify the pricing page has clear, structured pricing tables
  • Check if the page uses JavaScript rendering (may require Sphere integration)
  • Manual review and editing of the YAML may be needed for complex layouts

​
Type Misclassification

Problem: Features are classified with incorrect types Solutions:
  • Review the generated YAML and manually correct types
  • Update descriptions to provide clearer context for the LLM
  • Use the Pricing2Yaml specification as a guide for correct classifications

​
Pricing Discrepancies

Problem: Extracted prices don’t match the website Solutions:
  • Check if the page shows region-specific pricing
  • Verify the currency in the YAML matches the displayed currency
  • Use refresh: true to re-extract in case of cached stale data

​
Network or API Errors

Problem: Extraction fails with timeout or API errors Solutions:
  • Check A-MINT service logs: docker-compose logs a-mint-api
  • Verify OpenAI API key is valid and has available quota
  • Ensure the target URL is publicly accessible

​
Programmatic Usage

If you’re calling the MCP server programmatically, here’s an example tool call: 
from mcp import ClientSession

async with ClientSession(server_params) as session:
    result = await session.call_tool(
        "iPricing",
        {
            "pricing_url": "https://buffer.com/pricing",
            "refresh": False
        }
    )
    pricing_yaml = result["pricing_yaml"]

​
Next Steps

Optimization Queries

Learn how to find optimal plans using filters and constraints

Pricing Models

Deep dive into the complete Pricing2Yaml schema

Build docs developers (and LLMs) love

Get started for freeTalk to us