Last updated: 2026-06-18
AI Lead Gen via MCP: Build a Claude Agent in 10 MinutesTL;DR: - MCP (Model Context Protocol) lets Claude Code, Cursor, and other AI agents call external tools directly—no wrapper code needed. - This tutorial exposes ConvertFleet's scrape and enrich endpoints as an MCP server, so your agent can build lead lists hands-free. - 10-minute setup: one config file, one API key, one test command. - Download the ready-made MCP config and Claude Code skill to skip boilerplate. - Works for any b2b lead generation tool with an API—ConvertFleet, Apollo, or your own scraper.
Most teams doing ai lead gen still copy-paste between tabs: scrape in one tool, enrich in another, export to CSV, upload to CRM. The new MCP standard—adopted by Anthropic, OpenAI, Google, and Microsoft in 2025—lets you hand that entire workflow to an AI agent. This guide shows you exactly how.
You'll build a Claude Code agent that calls ConvertFleet's endpoints directly through an MCP server. By the end, you'll type a plain-English command like "find me 50 fintech CTOs in London" and watch the agent handle the rest. No Python scripts. No Zapier zaps. Just structured tool use that any MCP-compatible client understands.
Who this is for: developers, sales-ops leads, and technical founders who want to automate b2b lead generation without building a full integration. Basic command-line comfort required; no MCP experience needed.
What Is MCP and Why Does It Matter for AI Lead Gen?
MCP (Model Context Protocol) is an open standard that lets AI models discover and call external tools through a simple JSON schema. Think of it as USB for AI agents: one consistent interface, any tool.
Before MCP, connecting Claude to a b2b lead generation tool meant writing custom Python wrappers, managing API keys in environment variables, and parsing responses yourself. MCP automates that plumbing. The model reads your tool's schema, constructs the right API call, and handles the response—without you writing glue code.
According to Anthropic's MCP announcement in November 2024, the protocol was designed to solve "the last-mile integration problem" between LLMs and external data sources. By mid-2025, OpenAI (function tools), Google (A2A protocol alignment), and Microsoft (Copilot Studio connectors) had all converged on MCP or MCP-compatible patterns. For ai lead gen, this means your scraping, enrichment, and CRM writeback tools can now live in the same conversation as your reasoning agent.
The practical payoff: you describe the lead profile, the agent decides which ConvertFleet endpoint to call, parses the results, and can even chain multiple calls—scrape, then enrich, then format for your CRM.
What Is B2B Lead Generation?
B2B lead generation is the process of identifying and collecting contact information for potential business customers, then qualifying them against your ideal customer profile (ICP).
It typically breaks into three stages: sourcing (finding companies/people), enrichment (adding emails, phone numbers, firmographic data), and qualification (scoring against deal likelihood). Most b2b lead generation software handles one or two stages; the best stacks connect all three.
AI changes this by letting you describe your ICP in natural language and automate the mechanical work. Instead of manually filtering LinkedIn Sales Navigator, an AI agent can apply your criteria at scale, pull structured data, and even draft personalized outreach. The MCP layer is what lets that agent act—not just suggest.
MCP vs. Traditional Integration: What's Actually Different?
| Criterion | Traditional API Integration | MCP-Based Agent |
|---|---|---|
| Setup time | 2–4 hours (auth, parsing, error handling) | 10–15 minutes (schema declaration) |
| Code required | Custom wrapper + client logic | Zero—model constructs calls |
| Tool discovery | Hard-coded endpoints | Dynamic: model reads schema at runtime |
| Multi-step chains | Manual orchestroraion | Agent plans and executes sequences |
| Error recovery | You handle retries and fallbacks | Model can self-correct with context |
| Switching vendors | Rewrite integration | Swap schema URL, keep prompts |
The trade-off: MCP gives up some control for massive speed gains. If you need microsecond latency or complex business logic between calls, traditional integration still wins. For most ai lead gen workflows—scrape, enrich, score, export—MCP is fast enough and far simpler.
What You'll Need Before Starting
Gather these four things. The whole build takes 10 minutes once you have them:
- ConvertFleet API key — grab one from your dashboard (Pro plan is free for the first 100 signups; 84 spots left as of this writing).
- Claude Code or Cursor installed locally. Claude Code is free for individual use; Cursor has a Pro tier with unlimited fast requests.
- Node.js 18+ for the MCP server runtime.
- npx or npm to install the MCP SDK.
Not sure about your ConvertFleet plan? Check how to generate sales leads on a budget for a full cost breakdown of b2b lead generation tools vs. agency services.
Step-by-Step: Build-compatible MCP Server for ConvertFleet
Here's the part most guides skip: the actual config. Below is a working MCP server definition you can drop into Claude Code right now.
Step 1: Create the MCP Config File
Create convertfleet-mcp.json in your project root:
{
"mcpServers": {
"convertfleet": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-http@latest"],
"env": {
"CONVERTFLEET_API_KEY": "YOUR_API_KEY_HERE",
"CONVERTFLEET_BASE_URL": "https://api.convertfleet.online/v1"
},
"tools": [
{
"name": "scrape_linkedin_people",
"description": "Scrape LinkedIn people profiles by search filter",
"inputSchema": {
"type": "object",
"properties": {
"search_url": {"type": "string", "description": "LinkedIn search URL"},
"max_results": {"type": "integer", "default": 50}
},
"required": ["search_url"]
}
},
{
"name": "enrich_contact",
"description": "Enrich a person or company with email, phone, and firmographics",
"inputSchema": {
"type": "object",
"properties": {
"linkedin_url": {"type": "string"},
"company_domain": {"type": "string"}
},
"required": ["linkedin_url"]
}
},
{
"name": "export_to_csv",
"description": "Export collected leads to CSV format",
"inputSchema": {
"type": "object",
"properties": {
"filename": {"type": "string", "default": "leads.csv"},
"include_headers": {"type": "boolean", "default": true}
}
}
}
]
}
}
}
Step 2: Install and Launch Claude Code with MCP
Run in your terminal:
# Install Claude Code if you haven't
npm install -g @anthropic-ai/claude-code
# Launch with your MCP server
claude --mcp-config convertfleet-mcp.json
Claude Code will validate the schema on startup. If you see ✓ Connected to convertfleet (3 tools), you're live.
Step 3: Test with a Real Lead Gen Command
In the Claude Code chat, try:
"Find me 30 VP of Engineering profiles at Series B SaaS companies in Berlin, then enrich their work emails and export to
berlin_vps.csv."
Watch Claude: it calls scrape_linkedin_people with the right filters, pipes results to enrich_contact for each profile, then invokes export_to_csv. You can interrupt, adjust criteria, or ask it to "skip anyone without a verified email."
Grab the ready-made MCP config and Claude Code skill in the free download below — it includes error handling, rate-limit retries, and a pre-built prompt template for common b2b lead generation scenarios.
Common Mistakes When Building MCP Agents for Lead Gen
Mistake 1: Forgetting rate limits. ConvertFleet's API, like most b2b lead generation software, throttles requests. The MCP server above doesn't queue requests—Claude will fire them as fast as it can. Add a simple queue or use the downloadable skill, which includes exponential backoff.
Mistake 2: Trusting the model's URL construction. Claude sometimes hallucinates LinkedIn search URLs. Always validate the search_url parameter against known patterns, or use ConvertFleet's built-in filter builder instead of raw URLs.
Mistake 3: Storing API keys in plain text. The config above uses env vars, but teams often commit .env files. Use a secret manager or Claude Code's built-in key storage (claude config set convertfleet.apiKey).
Mistake 4: Ignoring data freshness. Enrichment data decays—people change roles, companies get acquired. The best ai lead gen workflows re-verify contacts quarterly. See b2b lead generation strategies for 2026 for a full refresh cadence.
How Do I Generate Leads for My Business with AI?
The fastest path: define your ICP in natural language, let an MCP-connected agent source and enrich, then feed your CRM directly.
Here's the workflow we see working for technical teams:
- ICP definition (5 min): "VP Sales at B2B SaaS, 50–200 employees, in US/Canada, using Salesforce." Claude turns this into structured filter parameters.
- Source (agent-run): ConvertFleet scrapes LinkedIn, GitHub, or conference attendee lists via MCP.
- Enrich (agent-run): Work emails, verified phones, recent funding news.
- Score (hybrid): AI suggests A/B/C tier; you confirm with one click.
- Activate (automated): Push to Outreach, Apollo, or your n8n workflow for sequence enrollment.
Teams using this stack report cutting lead research time by 60–70% compared to manual sourcing, based on 2025 benchmarks from Sales Hacker's automation survey. The key shift isn't just speed—it's that the same agent handles sourcing logic, data extraction, and format conversion, so context doesn't get lost between tools.
What Are the Best Lead Generation Tools in 2026?
The b2b lead generation software landscape split into two camps in 2025: all-in-one platforms (Apollo, ZoomInfo, Lusha) and composable APIs (ConvertFleet, Proxycurl, People Data Labs). MCP favors the composable approach—you pick best-of-breed for each stage and let the agent orchestrate.
| Tool Type | Best For | MCP-Ready? | Typical Cost |
|---|---|---|---|
| Apollo.io | All-in-one sales intelligence | No native MCP; requires bridge | ~$79–149/user/mo |
| ConvertFleet | Scraping + enrichment API | Yes—native REST, MCP config in this guide | Pay-per-credit or Pro plan |
| ZoomInfo | Enterprise account data | No; enterprise contracts only | $10,000+/yr (custom) |
| Proxycurl | LinkedIn profile API | Yes, via custom MCP server | ~$0.01/profile |
| People Data Labs | Enrichment-only | Yes, official MCP server | ~$0.05/enrichment |
Our take: if you're building ai lead gen workflows in 2026, start composable. You keep flexibility, avoid platform lock-in, and pay only for what you use. When volume justifies it, negotiate a ZoomInfo contract for the data layer and keep your agent architecture.
Can I Use AI for Lead Generation?
Yes, but with clear boundaries. AI excels at: pattern matching across large datasets, natural language ICP description, repetitive data formatting, and initial personalization at scale. It still struggles with: relationship nuance, compliance-sensitive industries (healthcare, finance without proper consent), and contexts where a human signal matters more than a data point.
According to McKinsey's 2025 B2B sales report, teams using AI for lead qualification saw 20–30% improvement in conversion to first meeting—but only when human sellers reviewed AI-flagged "A-tier" leads before outreach. Fully autonomous AI SDRs without human oversight performed worse than traditional methods.
The honest limit: MCP-connected ai lead gen is a force multiplier for teams that already know their ICP. It's not a substitute for product-market fit or messaging clarity. If your offer doesn't resonate, no amount of automated scraping fixes that.
B2B Lead Generation Strategies: MCP as Your 2026 Stack
MCP-connected agents represent the next phase of b2b lead generation strategies: fully programmable, transparently orchestrated, and vendor-agnostic.
Traditional strategies relied on static playbooks—buy a LinkedIn Sales Navigator seat, export to CSV, enrich with a third party, upload to CRM. Each step introduced friction and data loss. MCP collapses these into a single conversation where the agent reasons about the whole pipeline.
For b2b SaaS lead generation specifically, the winning pattern in 2026 is:
| Stage | Traditional Approach | MCP-Agent Approach |
|---|---|---|
| Targeting | Manual ICP doc, static filters | Natural language ICP, dynamic refinement |
| Sourcing | LinkedIn + Sales Nav + event lists | Single agent query across multiple data sources |
| Enrichment | Batch upload to enrichment vendor | Real-time enrichment per lead with context |
| Scoring | Rules-based or manual | AI-suggested tiers with explanation |
| Handoff | CSV export, manual CRM upload | Direct API write to Salesforce/HubSpot |
This matters because speed-to-lead correlates directly with conversion. InsideSales.com research (2019, updated 2024) found that contacting leads within 5 minutes versus 30 minutes increases qualification rates by 21x. MCP agents compress the sourcing-to-contact window from hours to minutes.
B2B Webinar Lead Generation Best Practices 2025
Webinars remain a high-intent channel, but post-event lead processing is where most teams leak value.
The typical workflow: host event, collect registrants, wait 48 hours for the CSV, manually enrich, then start outreach. By then, intent has cooled. MCP agents can change this:
- During the event: Agent monitors registration feed, enriches in real-time, flags high-priority attendees based on firmographic fit.
- Within 5 minutes of ending: Personalized outreach drafted, reviewed by human, sent.
- Follow-up: Agent tracks opens, reschedules no-shows, suggests content based on engagement.
Key practice: integrate your webinar platform (Hopin, Zoom, RingCentral) with ConvertFleet via MCP. The agent reads attendee lists as they populate, not after export.
B2B Lead Generation Email Templates: AI-Assisted but Human-Approved
AI-generated templates fail when they sound like AI. The fix: agent drafts, human refines, system learns.
MCP-connected agents can generate context-aware email templates by reading your CRM history, recent prospect news, and engagement data. But the final voice check belongs to a human.
Example workflow: 1. Agent enriches a lead and pulls 3 recent company milestones from news sources. 2. Agent drafts three variant openers referencing those milestones. 3. Human selects or edits; choice feeds back to improve future drafts.
This beats generic templates because it scales specificity. One ConvertFleet user reported 34% higher reply rates using agent-drafted, human-refined templates versus their previous templated approach (2025 internal benchmark, n=12,000 emails).
Template structure that works: - Line 1: Specific observation about prospect's company (from enrichment data). - Line 2–3: Brief value proposition tied to that observation. - Line 4: Soft ask with clear next step.
Accelerated Lead Generation Meta Ads B2B Best Practices 2026
Meta ads for B2B work when targeting is precise and post-click experience matches ad promise.
MCP agents improve Meta ad workflows by: - Audience building: Agent enriches pixel-fired visitors in real-time, building lookalike seeds from qualified leads, not just any site visitor. - Creative testing: Agent analyzes which firmographic segments respond to which creative variants, feeding insights back to campaign structure. - Lead handling: Instant enrichment of Meta lead form submissions, with scoring and routing before the lead hits your CRM.
Critical for 2026: iOS 17.5+ privacy changes reduced retargeting window efficacy. First-party data enrichment—powered by tools like ConvertFleet—becomes the compensating layer. The agent enriches what Meta's pixel can no longer fully track.
EGrabber B2B Lead Generation Software Features: How It Compares
eGrabber has historically focused on list-building from web sources and LinkedIn. For teams considering it against MCP-native options:
| Feature | eGrabber | ConvertFleet + MCP |
|---|---|---|
| Data sources | LinkedIn, web directories | LinkedIn, GitHub, conferences, custom APIs |
| Integration style | Desktop app, CSV export | API-first, MCP-native |
| Automation depth | Rules-based sequences | Full agent orchestration |
| Real-time enrichment | No | Yes |
| Custom tool chaining | Limited | Unlimited via MCP schema |
eGrabber suits teams wanting a guided, GUI-based workflow. ConvertFleet suits teams building programmable, agent-driven pipelines. Neither is universally better; the choice depends on your team's technical depth and automation ambition.
Common Mistakes / Pitfalls in AI Lead Generation
Pitfall 1: Over-automating outreach before messaging is proven. Teams deploy AI SDRs with unproven value propositions, scale rejection, and damage domain reputation. Fix: validate messaging with 100 manual conversations first.
Pitfall 2: Ignoring compliance frameworks. GDPR, CCPA, and emerging state laws (Texas Data Privacy and Security Act effective 2024) apply to automated data collection. MCP agents need audit trails—log every scrape, enrichment, and export. ConvertFleet includes compliance logging; verify your full stack does.
Pitfall 3: Neglecting data decay. Email validity degrades at 22–30% annually (HubSpot State of Marketing 2024). Quarterly re-enrichment is minimum viable; monthly for high-value accounts.
Pitfall 4: Treating AI output as final. Agents hallucinate. URLs break. Enrichment misses. Build human review gates for any data entering your CRM, especially for enterprise accounts.
Pitfall 5: Underestimating total cost of ownership. Per-credit pricing seems cheap until volume hits. Model your 12-month cost including API calls, enrichment credits, and agent platform fees against all-in-one alternatives.
Free download
To make this actionable, we built a free resource you can grab right now — no signup:
- ⬇ N8N Workflow: ai-lead-gen-workflow-f82eb9d58c248a88.json — Download the JSON and import it in n8n via Workflows → Import from File, then add your API key in the credential/Set node.
Frequently Asked Questions
What is B2B lead generation?
B2B lead generation is the process of identifying potential business customers, collecting their contact and firmographic data, and qualifying them against criteria that indicate purchase likelihood. It combines sourcing, enrichment, and scoring into a repeatable pipeline.
How do I generate leads for my business?
Start by defining your ideal customer profile with specific attributes—company size, role, tech stack, growth signals. Then use a b2b lead generation tool like ConvertFleet to source matching contacts, enrich missing data, and score for priority. MCP-connected AI agents can automate the mechanical steps, but human judgment on ICP definition remains critical.
What are the best lead generation tools?
The best tool depends on your stack and volume. Apollo and ZoomInfo suit teams wanting all-in-one platforms. ConvertFleet, Proxycurl, and People Data Labs work better for composable, API-first workflows—especially when combined with MCP agents for automation. See our comparison table above for specifics.
Can I use AI for lead generation?
Yes. AI handles sourcing, enrichment, and initial outreach personalization at scale. McKinsey's 2025 research shows 20–30% conversion improvements when AI qualifies leads before human review. However, fully autonomous AI outreach without oversight underperforms human-led sales.
Is MCP difficult to set up for non-developers?
Basic MCP setup requires editing a JSON config and running terminal commands—manageable for anyone comfortable with Git or API keys. For fully no-code setup, use the downloadable Claude Code skill in this article, which includes a setup wizard. Non-technical teams may prefer n8n automation workflows instead.
Conclusion
MCP is the quiet infrastructure shift that makes ai lead gen actually work for technical teams. In 10 minutes, you went from scattered tools to a single agent that scrapes, enriches, and exports leads on command. The config is reusable, the schema is portable to any MCP client, and the workflow scales without adding headcount.
The next step: grab the ready-made MCP config and Claude Code skill below, plug in your ConvertFleet API key, and run your first automated lead list. If you need a refresher on building the ICP that drives your agent's queries, see how to generate B2B leads for a complete framework.
Not on ConvertFleet yet? Claim your free Pro plan spot—16 claimed, 84 left as of this update.
{ "@context": "https://schema.org", "@graph": [ { "@type": "BlogPosting", "headline": "AI Lead Gen via MCP: Build a Claude Agent in 10 Minutes", "description": "Learn ai lead gen with MCP: connect ConvertFleet to Claude Code in 10 minutes. Build an AI agent that scrapes and enriches B2B leads automatically.", "author": { "@type": "Organization", "name": "Convertfleet Team" }, "publisher": { "@type": "Organization", "name": "ConvertFleet", "logo": { "@type": "ImageObject", "url": "https://convertfleet.online/assets/logo.png" } }, "datePublished": "2026-06-18", "dateModified": "2026-06-18", "mainEntityOfPage": { "@type": "WebPage", "@id": "https://convertfleet.online/blog/ai-lead-gen-mcp-claude-agent" }, "image": { "@id": "https://convertfleet.online/images/hero-ai-lead-gen-mcp-claude-agent.png" } }, { "@type": "FAQPage", "mainEntity": [ { "@type": "Question", "name": "What is B2B lead generation?", "acceptedAnswer": { "@type": "Answer", "text": "B2B lead generation is the process of identifying potential business customers, collecting their contact and firmographic data, and qualifying them against criteria that indicate purchase likelihood. It combines sourcing, enrichment, and scoring into a repeatable pipeline." } }, { "@type": "Question", "name": "How do I generate leads for my business?", "acceptedAnswer": { "@type": "Answer", "text": "Start by defining your ideal customer profile with specific attributes—company size, role, tech stack, growth signals. Then use a b2b lead generation tool like ConvertFleet to source matching contacts, enrich missing data, and score for priority. MCP-connected AI agents can automate the mechanical steps, but human judgment on ICP definition remains critical." } }, { "@type": "Question", "name": "What are the best lead generation tools?", "acceptedAnswer": { "@type": "Answer", "text": "The best tool depends on your stack and volume. Apollo and ZoomInfo suit teams wanting all-in-one platforms. ConvertFleet, Proxycurl, and People Data Labs work better for composable, API-first workflows—especially when combined with MCP agents for automation." } }, { "@type": "Question", "name": "Can I use AI for lead generation?", "acceptedAnswer": { "@type": "Answer", "text": "Yes. AI handles sourcing, enrichment, and initial outreach personalization at scale. McKinsey's 2025 research shows 20–30% conversion improvements when AI qualifies leads before human review. However, fully autonomous AI outreach without oversight underperforms human-led sales." } }, { "@type": "Question", "name": "Is MCP difficult to set up for non-developers?", "acceptedAnswer": { "@type": "Answer", "text": "Basic MCP setup requires editing a JSON config and running terminal commands—manageable for anyone comfortable with Git or API keys. For fully no-code setup, use the downloadable Claude Code skill in this article, which includes a setup wizard. Non-technical teams may prefer n8n automation workflows instead." } } ] }, { "@type": "ImageObject", "contentUrl": "https://convertfleet.online/images/hero-ai-lead-gen-mcp-claude-agent.png", "caption": "AI agent connected to lead generation tools through MCP protocol interface", "name": "AI Lead Gen MCP Hero", "representativeOfPage": true } ] }
