Hey everyone! If you’ve been following the AI space lately, you’ve probably heard a lot of buzz about “AI agents” and something called MCP. But what does it all mean, and more importantly, how can you actually use this stuff to automate your work? Today, I’m going to break down everything you need to know about automation using AI agents and the MCP Toolkit—no PhD required.
What Exactly Is an AI Agent?
Before we dive into the technical bits, let’s get clear on what we’re talking about. An AI agent isn’t just a chatbot that answers questions. It’s an autonomous system that can reason, plan, and take action on your behalf. Think of it as the difference between asking someone for directions versus having a personal assistant who books your flight, reserves your hotel, and sends you the itinerary.
AI agents can:
- Break down complex tasks into manageable steps
- Use external tools and services to accomplish goals
- Make decisions based on context and feedback
- Learn from interactions to improve over time
The key word here is autonomy. These systems don’t just respond—they act. And that’s where the real automation magic happens.
Enter MCP: The Model Context Protocol
Now, here’s where things get interesting. Back in November 2024, Anthropic (the folks behind Claude) introduced the Model Context Protocol (MCP)—an open standard that’s basically become the USB-C of AI integration. Just like USB-C standardized how we connect devices, MCP standardizes how AI models connect to external tools and data sources.
MCP is an open protocol that enables seamless integration between large language model (LLM) applications and external data sources and tools. Before MCP, connecting an AI to your database, your calendar, your code repository, or any other service required custom integration work for each connection. It was messy, fragmented, and honestly kind of a nightmare.
With MCP, you get:
- Standardized connections – One protocol to rule them all
- Tool discovery – AI can automatically find and use available tools
- Secure communication – Built-in safety mechanisms
- Model-agnostic design – Works across different AI providers
Why 2025-2026 Is the Tipping Point
If 2025 was the year of MCP adoption, 2026 is shaping up to be the year of expansion. Nearly every major player has jumped on board. OpenAI, Google DeepMind, and of course Anthropic all support MCP now. Tools like Figma, Notion, Linear, Atlassian, Zapier, Stripe, PayPal, MongoDB, and countless others have built MCP servers that work seamlessly together.
The protocol has gotten so significant that Anthropic moved MCP into a newly founded open source foundation under the Linux Foundation called the Agentic AI Foundation (AAIF). That’s a big deal—it signals that MCP isn’t going anywhere and that the industry is committed to maintaining it as a truly open standard.
The MCP Toolkit Ecosystem
So how do you actually use all this? Let’s talk about the tools available to you.
AI Toolkit for Visual Studio Code
Microsoft has released the AI Toolkit for Visual Studio Code, which is probably the most accessible way to start building AI agents with MCP support. The extension includes:
- Model Discovery – Browse and access AI models from GitHub, ONNX, Ollama, OpenAI, Anthropic, and Google
- Prompt Engineering Tools – Create and optimize prompts with iterative improvement
- Agent Builder – A visual interface for creating and optimizing agents
- MCP Server Integration – Connect your agents to external tools through MCP servers
The Agent Builder (formerly Prompt Builder) lets you create, iterate, and optimize agents from prompt engineering to tool integration—all in one seamless workflow. It’s genuinely impressive how much they’ve lowered the barrier to entry.
Docker MCP Toolkit
Docker has integrated MCP Toolkit directly into Docker Desktop. This lets you host and manage MCP servers as modular tool endpoints running inside Docker containers. If you’re already comfortable with Docker, this is a natural fit. You can find it by navigating to the MCP Toolkit tab in Docker Desktop.
VS Code Native MCP Support
VS Code now has built-in MCP support in Agent Mode with GitHub Copilot. Once you’ve added an MCP server, you can use the tools it provides directly in chat. The agent can perform browser actions, connect to your chat and note-taking apps, interact with databases, access your design system, pull issues from GitHub—the list goes on.
Real-World Automation Use Cases
Okay, theory is great, but let’s talk about what you can actually do with this stuff. Here are some practical automation workflows that people are building right now:
1. Meeting and Calendar Automation
Problem: Scheduling meetings across multiple calendars and rooms wastes time and leads to endless back-and-forth messages. With MCP, an AI agent can check availability across all participants, find suitable times, book the meeting, and send invites—all from a simple natural language request. Meetings get scheduled in seconds, not minutes.
2. Document Creation from Conversations
By connecting Slack and Google Docs MCP servers, an AI agent can automatically create well-structured documents from conversations—things like postmortems, meeting summaries, or strategy documents. And because MCP works across tools, you can adapt this to Notion, Confluence, or wherever your team works.
3. Sales and Customer Success Prep
Sales and CS teams often waste time digging through emails and CRM notes before client meetings. An agent using HubSpot and Gmail MCP servers can automatically compile relevant customer history, recent interactions, and talking points before each meeting. Better meetings, faster follow-ups, less prep time.
4. HR Onboarding Automation
When a new employee is hired, the system can automatically create an entry in Notion’s onboarding database, populate it with the standard checklist, and send a personalized welcome email with a direct link. The HR team saves administrative time, and new hires get a seamless start.
5. Customer Support Triage
Using a help desk MCP server like Zendesk, incoming support tickets get automatically analyzed for urgency and topic, categorized appropriately, and either routed to the right agent or auto-responded with relevant knowledge base articles. Critical issues get prioritized without human intervention.
6. Code Management and Development
By integrating GitHub MCP with an AI agent, you can use natural language to update repositories, create branches, commit changes, and manage pull requests. It automates routine code management, reduces manual errors, and lets developers focus on actual coding.
Building Your First MCP Agent
Ready to try this yourself? Here’s a high-level overview of how to get started:
Choose Your Framework
There are now at least 12 major agent SDKs with MCP support. Popular options include:
- OpenAI Agents SDK – Great for building assistants on OpenAI’s platform, with built-in MCP support via MCPServerStdio and MCPServerSse classes
- mcp-agent by LastMile AI – Focuses on simple, composable patterns rather than complex architectures
- LangChain MCP Adapters – A lightweight wrapper that converts MCP tools into LangChain tools for use with LangGraph agents
- Claude Agent SDK – Security-first approach for production environments
- CrewAI – Excellent for multi-agent workflows
Start Simple
A common starter project is creating an MCP server that fetches weather data for a specified location. FastMCP is a Python library that makes implementing MCP servers straightforward. Once you’ve got the basics down, you can build more complex workflows.
Use the Visual Tools
If you prefer a more visual approach, OpenAI’s Agent Builder offers drag-and-drop, node-based agent building with MCP server support. Microsoft’s AI Toolkit for VS Code also provides scaffolding to help you build new MCP servers with a basic protocol implementation you can customize.
Security Considerations
Before you go wild automating everything, let’s talk about safety. In April 2025, security researchers identified several outstanding issues with MCP, including:
- Prompt injection vulnerabilities
- Tool permission issues where combining tools can exfiltrate files
- Lookalike tools that can silently replace trusted ones
The official recommendation is that there should always be a human in the loop with the ability to deny tool invocations. Don’t give your AI agents free rein over sensitive systems without proper oversight and permission controls.
What’s Coming Next
The trajectory here is clear. 2026 is the year AI goes truly agentic—models that can reason, act, and help you across multiple tools in real-time. MCP is the backbone making this possible.
We’re likely to see:
- More parallel execution – Running multiple tasks in the background while starting new ones
- Better MCP server management – As organizations deploy more active MCP servers, management tools will become crucial
- Deeper integration – MCP becoming the default infrastructure for contextual AI
- Cross-server workflows – More sophisticated automations coordinating multiple MCP servers
Getting Started Today
If you’re excited about this (and honestly, you should be), here’s my recommendation for getting started:
- Install the AI Toolkit for VS Code – It’s the lowest-friction way to experiment
- Check out the GitHub MCP Registry – It’s the new home base for discovering MCP servers
- Start with a simple use case – File organization or calendar automation are good first projects
- Join the community – The MCP ecosystem is growing fast, and there’s tons of shared knowledge
The tools and development practices are still maturing, but that’s also what makes this an exciting time to get involved. You’re not just learning a new tool—you’re getting in on the ground floor of how we’ll all be working with AI in the years to come.
The future of automation isn’t about replacing humans—it’s about giving us superpowers. MCP and AI agents are the infrastructure that makes those superpowers possible. Now go build something cool.
Leave a Reply