LeadFlow OS

Local Business Intelligence Automation Kit for Apify, n8n, and MCP-agent workflows.

What it does

LeadFlow OS turns local-business records into a workflow-ready lead pipeline:

Google Maps-style business records
-> clean and normalize
-> remove duplicates
-> score each lead
-> split outputs into useful datasets
-> send A/B leads to n8n, Sheets, CRM, Telegram, or an AI agent through Apify MCP

This Actor is intentionally not another raw scraper. The product gap is the layer after scraping: the cleaning, scoring, routing, reporting, and automation that makes scraped data useful.

Why this exists

The elephant in the room is Google Maps/local-business scraping demand. Huge existing Actors already collect raw places data. LeadFlow OS does not fight that elephant. It sells the elephant a dashboard.

Use any Google Maps-style export, paste the records here, and get CRM/n8n/MCP-ready rows back.

MCP-ready positioning

Apify's hosted MCP server at https://mcp.apify.com can expose Actors as tools to AI clients. LeadFlow OS is prepared for that flow with:

• a clear input schema for business records and filters
• an output schema that points agents to the default dataset and run summary
• explainable lead scores and reason codes
• a mcpAgentHint field reminding agents to ask before writing to CRM or contacting businesses

Example agent flow:

AI agent receives: "Find the best elephant-themed product shops from this Apify dataset."
-> calls LeadFlow OS through Apify MCP
-> receives scored rows
-> summarizes A-grade leads
-> asks before pushing anything to a CRM or outreach system

MVP scope

v0.1.1 is post-processing mode. It accepts already-collected business records and turns them into clean, scored, workflow-ready output.

Included now:

• record normalization
• website/domain normalization
• deduplication by domain or name/address
• explainable lead scoring
• filters for rating, reviews, and required website
• multiple named datasets
• n8n template for high-score leads to Google Sheets plus Telegram
• output schema for better Apify Console/API/MCP discovery

Not included yet:

• built-in Google Maps crawling
• contact-page/email enrichment
• review summarization
• direct CRM writes

Those are v0.2+ candidates.

Example input

{
  "businesses": [
    {
      "name": "Elephant & Co. Toy Store",
      "address": "Kochi, Kerala",
      "phone": "+91 99999 99999",
      "website": "elephant-toys.example",
      "rating": 4.7,
      "reviewCount": 183,
      "category": "Toy store"
    },
    {
      "name": "Product Nest Gifts",
      "address": "Kochi, Kerala",
      "rating": 3.6,
      "reviewCount": 42,
      "category": "Gift shop"
    }
  ],
  "leadScoringProfile": "local_services",
  "outputMode": "mcp_agent_ready"
}

Main output fields

Field	Meaning
name	Business name
category	Business category
address	Address/location text
phone	Phone number if present
website	Normalized website URL
domain	Website domain for dedupe/routing
rating	Numeric rating
reviewCount	Number of reviews
leadScore	0-100 score
leadGrade	A/B/C/D grade
scoreReasons	Explainable reason codes
workflowStatus	dataset_only, n8n_ready, crm_ready, or mcp_agent_ready
suggestedNextAction	Recommended workflow action
mcpAgentHint	Agent safety hint in MCP-agent mode

Output datasets

Dataset alias	Purpose
default	Outreach-ready records for CRM/n8n/MCP agents
businesses_clean	Normalized business records
lead_scores	Score, grade, and reason codes
failed_locations	Skipped or invalid records
duplicates	Duplicate records
audit_debug	Run summary and non-sensitive diagnostics

The Actor also stores LEADFLOW_SUMMARY in the default key-value store.

Lead scoring v0.1.1

The scoring is simple and explainable:

• business name present
• phone present
• website present
• strong rating or low-rating opportunity
• review volume
• category present
• optional boosts for agency/local-services profiles

Grades:

• A: 75+
• B: 55-74
• C: 35-54
• D: below 35

n8n template

Included template:

templates/n8n/leadflow-os-google-sheets-telegram.json

Workflow shape:

Manual/Cron trigger
-> Run LeadFlow OS Actor through Apify API
-> Read default dataset
-> Filter A/B leads
-> Append to Google Sheets
-> Send Telegram digest

Pricing / Cost estimation

LeadFlow OS uses Apify's pay-per-event pricing model.

Event	Price
Actor Start	$0.00005 per run
Outreach-ready lead	$0.003 per lead ($3 per 1,000)

Only records that pass filtering, deduplication, and scoring - and land in the default dataset - are billed as an "Outreach-ready lead." The run summary's outreachReadyRecords count is exactly the number of billable leads for that run, so what you see in the summary is what you pay for. Records skipped into failed_locations or duplicates are never billed.

Permission and safety model

• Basic mode is dataset-only/post-processing. No workspace writes.
• MCP-agent mode returns structured lead intelligence, but does not contact businesses or write to CRM by itself.
• Advanced future connector modes should use least privilege and document any approval prompts clearly.

Local development

npm install
npm run check
npm start

Product note

LeadFlow OS is a productized automation layer around proven Apify demand. The sellable result is not "I scraped more rows." The sellable result is "your leads are cleaned, scored, routed, and ready for action."