Claude SkillMCPAI AutomationPythonNext.js

AI Lead Automation Hub

Turning repetitive lead processing into an AI-powered workflow using Claude Skills and MCP.

Run Interactive Demo View Workflow

Automation workflow ready

  1. Lead
  2. Claude Skill
  3. MCP
  4. External Tool
  5. Result

Manual lead processing is repetitive.

A typical lead needs someone to do all of this, every time:

  • Read the lead
  • Understand intent
  • Classify it
  • Write a response
  • Update the CRM
  • Create a follow-up
  • Track the result

  • Why automate it?

    The process is repetitive, structured, and happens more than once.

From AI that thinks to AI that acts.

The Skill tells Claude how to do the job. MCP gives Claude the tools to do it.

Claude

Reads, reasons and writes.

Skill

Inputs, steps, quality standards, output format.

MCP

Tools that let the AI reach external systems.

External tools

CRM, email, calendar. Planned integrations.

The workflow

Select a step to see what happens inside it.

Claude Skill

Defined by the Skill

Claude loads process-new-leads (SKILL.md). The Skill defines inputs, steps, quality standards and output format.

Live demo

Process a lead and watch the workflow run.

Demo mode — external MCP integrations are not connected.

  • Lead received
  • Input validated
  • Intent analyzed
  • Lead classified
  • MCP tool selected
  • CRM update simulated
  • Follow-up created (simulated)
  • Processing complete

The Claude Skill

A Skill is a SKILL.md file that tells Claude how to do one job well.

process-new-leads/SKILL.md

---

name: process-new-leads

description: Analyze, qualify, communicate with, and schedule follow-ups for new business leads.

---

# INPUTS

  • - Name
  • - Email
  • - Company
  • - Interest
  • - Message

# WORKFLOW

  • 1. Validate lead information
  • 2. Analyze intent
  • 3. Classify lead
  • 4. Generate follow-up
  • 5. Execute MCP actions
  • 6. Return processing report

# QUALITY STANDARDS

  • - Never invent missing information
  • - Keep communication professional
  • - Explain classification
  • - Report every action
  • - Never claim an action happened unless the tool confirms it

# OUTPUT

  • - Lead
  • - Classification
  • - Reason
  • - Action status
  • - Next action

MCP architecture

MCP acts as the bridge between the AI and external tools.

Claude
SKILL.md
MCP Server
get_new_leads
update_lead
create_followup
Real tool (CRM, email, calendar)
Planned integration. In this site, tool calls are simulated.
  • Claude reads the Skill and decides which tool fits each step.
  • The MCP server exposes a small set of tools with typed inputs.
  • The real tool does the work and confirms the result. The Skill only reports what a tool confirms.

MCP server code

A minimal Python server with two tools.

server.py
from mcp.server.fastmcp import FastMCP
mcp = FastMCP("Lead Automation")
@mcp.tool()
def update_lead(email: str, status: str):
    """Update the lead status."""
    return f"{email} updated to {status}"
@mcp.tool()
def create_followup(email: str, days: int):
    """Create a follow-up task."""
    return f"Follow-up created for {email} in {days} days"
FastMCP(...)
Creates the MCP server and names it.
@mcp.tool()
Registers a function so Claude can call it.
Type hints
Tell Claude what each input must be.
Docstring
Claude reads it to decide when to use the tool.
return
The confirmation the Skill relies on. Here it is a plain string, not a real update.

Before vs after

Before: manual

  1. New leadthen
  2. Open CRMthen
  3. Read leadthen
  4. Analyzethen
  5. Write emailthen
  6. Create reminderthen
  7. Update CRM

7 steps

After: automated

  1. New leadthen
  2. Claude Skillthen
  3. MCPthen
  4. Automated actionsthen
  5. Result

5 steps

Test cases

The Skill is tested twice with different inputs. Outputs below come from the demo's local mock classifier.

Test 01

Completed
Input
Carlos Pérez, TechBol, AI Automation
Expected
High Intent
Output
HIGH INTENT
Reason
The lead explicitly requested information about AI automation for their company.

Test 02

Completed
Input
Maria López, Nova Systems, General information request
Expected
Medium Intent
Output
MEDIUM INTENT
Reason
The lead asked for general information without a specific need or timeline.

Tech stack

  • Claude

    The model that reads leads and reasons about intent.

  • Claude Skills

    SKILL.md files that package a repeatable process.

  • MCP

    Open protocol connecting AI to external tools.

  • Python

    Language of the example MCP server (FastMCP).

  • Next.js

    Framework for this presentation site.

  • TypeScript

    Typed code across the site.

  • Tailwind CSS

    Utility styling for the dark UI.

  • Vercel

    Target platform for deployment.