CrewAI MCP Lead Gen Agents

A multi-agent system that generates qualified B2B lead lists with enriched contact data and personalized outreach templates — powered by CrewAI, Google Gemini, and 5 production MCP servers from the Vinkius AI Gateway.

License: MIT Python 3.11+

The Problem with Manual Prospecting

Sales teams spend an average of 21% of their day researching prospects. An SDR manually checking LinkedIn, searching Crunchbase for funding data, running Hunter.io for email addresses, and then writing a personalized email for each lead — that is 2+ hours per 10 leads, and the quality degrades with fatigue.

This project automates the entire pipeline. Three AI agents handle company research, contact enrichment, and outreach personalization in a single run. You define your Ideal Customer Profile, and the system delivers a qualified lead list with verified emails and ready-to-send outreach templates.

The key differentiator is that the agents access live data through the Model Context Protocol (MCP) — they query Apollo.io for firmographics, Crunchbase for funding data, Hunter for verified emails, LinkedIn for professional profiles, and Exa AI for recent company signals. No stale data. No guessing.

How It Works

Phase 1 — Company Research and Qualification

The first agent acts as a senior sales researcher. It uses two MCP servers:

  • Apollo.io for company search, firmographics, employee count, revenue estimates, and technology stack detection
  • Crunchbase for funding rounds, investors, total funding raised, and acquisition history

Given your ICP (industry, company size, product fit), it finds 10 qualified companies, scores each for fit from 1-10, and provides reasoning for every score.

Phase 2 — Contact Enrichment

The second agent finds decision-makers at each qualified company. It uses two MCP servers:

  • Hunter for email discovery and verification with confidence scores
  • LinkedIn for professional profiles, current titles, seniority levels, and department information

It targets the specific role you define (VP of Engineering, Head of Product, CFO) and prioritizes email accuracy over speed.

Phase 3 — Personalized Outreach

The third agent creates outreach templates. It uses one MCP server:

  • Exa AI for semantic search of recent company news, blog posts, and hiring signals

Each email references something specific about the prospect's company — a recent funding round, a hiring trend, a technology adoption — connecting it to your value proposition. Emails are kept under 120 words with a single clear call to action.

Only 5 MCP Servers Required

This project is intentionally lean. You need only 5 MCP servers to run the full pipeline:

MCP Server Agent Purpose
apolloio-mcp Lead Researcher Company search, firmographics, tech stack
crunchbase-mcp Lead Researcher Funding data, investors, company profiles
hunter-mcp Contact Enricher Email discovery and verification
linkedin-mcp Contact Enricher Professional profiles, job titles
exa-ai-mcp Outreach Strategist Company news, signals, personalization data

All hosted on the Vinkius AI Gateway. Deploy each one in under a minute and get your SSE endpoint URL.

Want More Data Sources?

The Vinkius AI Gateway offers additional sales intelligence MCP servers you can add with a one-line config change:

Browse the full catalog of 2,600+ production-ready MCP servers at vinkius.com/en/categories.

Getting Started

Prerequisites

Installation

git clone https://github.com/vinkius-labs/crewai-mcp-lead-gen-agents.git
cd crewai-mcp-lead-gen-agents

python -m venv .venv
source .venv/bin/activate    # Linux/macOS
# .venv\Scripts\activate     # Windows

pip install -e .

Configuration

cp .env.example .env

Add your Gemini API key and 5 Vinkius MCP URLs.

Usage

# Validate configuration
lead-gen validate

# Generate leads for FinTech VP of Engineering
lead-gen prospect "FinTech" "VP of Engineering" \
  --size "50-500" \
  --product "API monitoring platform" \
  --value "reduce API downtime by 90%"

# Generate leads for SaaS Head of Product
lead-gen prospect "SaaS" "Head of Product" \
  --size "100-1000" \
  --product "user analytics platform" \
  --value "increase feature adoption by 3x"

Generated Output

The system produces a markdown report containing:

Section Content
Company Profiles 10 qualified companies with firmographics, fit scores, and reasoning
Decision-Maker Contacts Verified emails, LinkedIn profiles, titles, seniority
Outreach Templates 10 personalized emails with subject lines and CTAs
Summary Metrics Total leads, average fit score, email coverage rate

Technical Details

  • Framework: CrewAI with Flows and @CrewBase decorators
  • LLM: Google Gemini 2.0 Flash (free tier compatible)
  • State: Pydantic models — CompanyProfile, ContactProfile, OutreachMessage
  • MCP: Native CrewAI mcps= field with SSE transport to Vinkius AI Gateway
  • CLI: Typer with ICP-based arguments

FAQ

What is MCP?

The Model Context Protocol is an open standard for connecting AI systems to external tools. See modelcontextprotocol.io.

Can I push leads directly to my CRM?

Yes. Add the HubSpot, Salesforce, or Pipedrive MCP server and extend the outreach agent to push leads directly.

Is this GDPR compliant?

This tool discovers publicly available business contact information. However, compliance depends on your jurisdiction and how you use the data. Consult your legal team before using enriched contact data for outbound campaigns.

Contributing

Please read the Contributing Guide before submitting a pull request.

License

MIT — see LICENSE.

Built by Vinkius Labs with CrewAI and the Vinkius AI Gateway.