🤖 MCP + AI Agents: A Live Demo

See how Model Context Protocol transforms AI interactions with structured context

❌ Without MCP Limited

Manual copy-paste: You have to copy career data, metrics, and achievements into every conversation
Context loss: AI forgets previous conversations, you repeat yourself constantly
Inconsistent data: Different answers because you paste different info each time
Time-consuming: Every query requires finding and formatting data manually
Error-prone: Easy to miss important details or provide outdated information

✅ With MCP Powerful

Automatic context: AI instantly accesses your complete career history, metrics, and expertise
Structured data: Career achievements, quantifiable metrics, and timeline always available
Consistent answers: Same source of truth every time, no data drift
Smart queries: AI can search, filter, and extract specific information on demand
Always current: Update once in your files, AI sees changes immediately

🎮 Try It Live: Simulated MCP Queries

Click the buttons below to see what an AI agent can retrieve through MCP

👆 Click a button above to see MCP in action

Why MCP + AI Agents Matter

⚡

Speed

AI accesses your complete context in milliseconds, not minutes of manual prep

🎯

Accuracy

Structured data means consistent, factual answers based on your actual achievements

🔄

Reusability

Define your context once, use it across unlimited conversations and tools

🛡️

Control

You own and control exactly what data the AI can access through MCP

📚

Scalability

Works with any MCP-compatible AI client: Claude, custom tools, future apps

🔐

Privacy

Runs locally on your machine via stdio—no external servers or data leaks

Real-World Use Cases

💼 Executive Resume Generation

Scenario: You need to apply for a board position and want a tailored executive summary.

With MCP: Ask AI "Generate a board-focused executive summary emphasizing M&A and EBITDA impact" — it pulls your 7 acquisitions, OPEX reduction from 20% to <15%, and PE success story automatically.

Without MCP: You manually dig through old resumes, LinkedIn, and notes to find metrics, then copy-paste everything into the AI prompt.

🎤 Presentation Prep

Scenario: Speaking at a conference about scaling engineering teams.

With MCP: "Create talk outline highlighting team growth, engagement improvements, and retention" — AI instantly retrieves: 160→400+ scaling, 56%→75% engagement, >90% retention.

Without MCP: You search emails, old decks, and spreadsheets to reconstruct your achievements.

📝 Performance Review

Scenario: Writing annual self-assessment for your CEO and board.

With MCP: "Summarize my 2024 achievements with quantifiable metrics" — AI extracts all relevant data: cloud cost savings, NPS improvements, AI product launches, team growth.

Without MCP: You piece together achievements from Slack messages, Jira, and quarterly reports over several hours.

🤝 Networking & Introductions

Scenario: Investor asks for a quick bio before intro call.

With MCP: "Generate 3 bio variations: results-focused, leadership-focused, business-impact" — Done in 5 seconds with accurate metrics.

Without MCP: You write from scratch or repurpose old content that's potentially outdated.

🎓 Teaching & Mentoring

Scenario: Mentoring a new CTO on scaling engineering organizations.

With MCP: "Show me progression of team scaling decisions and outcomes across my career" — AI maps out your journey from 100-person to 400+ person org with lessons learned.

Without MCP: You try to remember which decisions happened at which company, missing context and timelines.

📊 M&A Due Diligence

Scenario: PE firm asks about your M&A integration experience.

With MCP: "Detail my 7 acquisitions with integration strategies and outcomes" — AI surfaces platform consolidation approaches, team integration, and ARR growth results.

Without MCP: You hunt through old PowerPoints and try to recall which metrics went with which acquisition.

🎯 Job Interview Prep

Scenario: Preparing for CTO interview at high-growth SaaS company.

With MCP: "Create STAR-format answers for scaling teams, cloud cost optimization, and AI adoption" — AI generates complete, metric-backed responses.

Without MCP: You create a prep doc manually, miss key metrics, and lack specific examples.

📰 Press & Media

Scenario: Journalist writing article about AI in enterprise engineering.

With MCP: "Generate talking points about AI adoption with concrete examples" — AI pulls your AI product launches, productivity gains, and center of excellence work.

Without MCP: You scramble to recall achievements and hope you don't misstate a metric in the interview.

💡 Strategic Planning

Scenario: Building 3-year technology roadmap for new role.

With MCP: "What were my biggest wins in platform modernization and cloud migration?" — AI shows you what worked: cost savings, performance improvements, customer impact.

Without MCP: You rely on memory of what worked vs what didn't, potentially repeating mistakes.

🎬 See MCP in Action: Visual Walkthrough

Watch how an AI agent uses MCP to answer questions in real-time

👤

Step 1: User Asks Question

User:

"What were my biggest team scaling achievements at Quorum?"

⏱️ Total time: ~100ms | 🔒 All data stays local | 🎯 100% accurate metrics

How It Works (Technical Overview)

1. Define Your Context (One Time)

            
// career.md - Your structured career data
## Quorum Software - CTO
- Scaled engineering from 160 to 400+
- Improved engagement: 56% → 75%
- Led 7 acquisitions
- Reduced OPEX: ~20% → <15%
            
          

2. Expose via MCP Server

            
// MCP server exposes resources and tools
resources: [
  "tomlacy://career/summary",
  "tomlacy://metrics/all",
  "tomlacy://expertise/areas"
]

tools: [
  query-career(keyword),
  generate-bio(context),
  extract-metrics(category)
]
            
          

3. AI Agent Accesses Context

            
// User asks Claude:
"Show me my team growth metrics"

// Claude uses MCP to call:
extract-metrics({ category: "Team Growth" })

// Returns instantly:
{
  metric: "Team size",
  value: "160 → 400+",
  context: "Quorum Software"
}
            
          

4. Get Instant, Accurate Results

AI generates responses using your exact data—no hallucinations, no outdated info, no manual copy-paste.

🛠️ Build Your Own MCP Server

Creating an MCP server for your own data is easier than you think. Here's a step-by-step guide.

MCP Server Architecture

              📱 AI Client (Claude Desktop, Custom Apps)
            
↕️

              MCP Protocol (stdio / HTTP)
            
↕️

              🔌 Your MCP Server (Node.js / Python)
            
↕️

                📄 Resources
(career.md)
              
                🔧 Tools
(search, extract)
              
                💡 Prompts
(templates)

Step 1: Install MCP SDK

              
# For Node.js
npm install @modelcontextprotocol/sdk

# For Python
pip install mcp

Step 2: Define Your Resources

Resources are read-only data sources (files, databases, APIs).

              
// Define what data you want to expose
const resources = [
  {
    uri: "mydata://resume",
    name: "My Resume",
    description: "Complete professional resume",
    mimeType: "text/markdown"
  },
  {
    uri: "mydata://projects",
    name: "Project Portfolio",
    description: "All completed projects with outcomes",
    mimeType: "application/json"
  }
];
              
            

Step 3: Create Tools for Queries

Tools allow AI to search, filter, and transform your data.

              
// Define tools the AI can use
const tools = [
  {
    name: "search-projects",
    description: "Search projects by technology or outcome",
    inputSchema: {
      type: "object",
      properties: {
        query: { type: "string" },
        technology: { type: "string" }
      }
    }
  }
];
              
            

Step 4: Implement Tool Handlers

Write the logic that executes when AI calls your tools.

              
async function handleToolCall(toolName, args) {
  if (toolName === "search-projects") {
    // Load your data
    const projects = await loadProjects();
    
    // Filter by query
    const results = projects.filter(p => 
      p.description.includes(args.query)
    );
    
    return { results };
  }
}
              
            

Step 5: Start the MCP Server

Connect via stdio for local use or HTTP for remote access.

              
import { Server } from '@modelcontextprotocol/sdk/server/index.js';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';

const server = new Server({ name: "my-mcp-server" });
const transport = new StdioServerTransport();

await server.connect(transport);
console.log("MCP Server running!");

Step 6: Connect to Claude Desktop

Add your server to Claude's configuration file.

              
// ~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "my-server": {
      "command": "node",
      "args": ["/path/to/your/server.js"]
    }
  }
}
              
            

💡 Tips & Best Practices

Start simple: Begin with one resource (like your resume) before adding tools
Use structured data: JSON and Markdown are easier for AI to parse than PDFs
Add descriptions: Clear descriptions help AI know when to use each resource/tool
Test locally first: Use stdio transport for local testing before deploying
Version your data: Keep career data in Git so you can track changes
Consider privacy: Only expose data you're comfortable with AI accessing
Monitor usage: Log which resources and tools get called to optimize

MCP Server Ideas by Profession

👨‍💼 Executives

Board presentations library
Quarterly metrics and KPIs
M&A transaction history
Strategic initiatives tracker

👨‍💻 Developers

GitHub repos and contributions
Technical blog posts
Stack Overflow answers
Project architecture docs

📊 Analysts

Analysis methodologies
Data models and schemas
Insight repository
Tool and technique catalog

🎨 Designers

Portfolio case studies
Design system guidelines
User research findings
A/B test results

📝 Writers

Published articles
Writing style guide
Research notes
Interview transcripts

🔬 Researchers

Publication database
Experiment protocols
Dataset catalog
Citations and references

Ready to Try the Real Thing?

Tom's actual MCP server is open source and ready to use with Claude Desktop or any MCP client.

📦 View MCP Server on GitHub 📖 Read Setup Guide