Generative AI Series
Vibe Coding (Agentic Coding) — MCP powers the Vibe (2/2)
In this blog, I am sharing some of my learnings, Vibe coding for the past few weeks, specifically in this blog, I will share how MCP Servers accelerated my vibe coding workflow.
In the first part of the blog, I covered some of my experiences vibe coding for the past few weeks. Please check that out here: Vibe Coding (Agentic Coding) — My experiences (1/2). I did mention MCP servers, but I thought MCP servers needed a separate blog, as they power the vibe coding.
As mentioned in the previous blog, Context is king to make Agentic AI work better and provide more reliable results. I’ve been experimenting with ways to make that flow smoother, faster, and honestly, just more fun. That’s where MCP servers have completely changed the game for me.
MCP stands for Model Context Protocol, and while that sounds like some enterprise spec sheet jargon, it’s basically a way to get your AI tools and dev assistants/agents to actually talk to each other. This blog is not really about MCP architecture, or how to install MCP or build a MCP server….maybe that is a topic for later, there is already enough on this topic. Please read it then before you go any further.
In this blog post, I will be sharing some of the top MCP servers I’ve been using to boost my vibe coding. These tools have been a game-changer for me, and I’m eager to share them with you. I won’t go into the technical details of how I set them up, as the documentation is all there for that. Instead, I’ll list them and discuss how they can help you improve your coding experience.
General
Context7: One of the must-haves, IMHO. This helps you solves the critical problem of providing version-specific documentation and code snippets directly into your LLM-powered IDE. It’s super smart and understands the context of your project. So, if you’re working with a specific version of programming language, framework, or platform, it’ll fetch the exact matching documentation, examples, and usage patterns. Context 7
GitHub MCP Server leads the ecosystem and serves as the gold standard for repository management. It provides comprehensive Git operations, issue tracking, pull request management, and seamless API integration. The server supports both GitHub.com and GitHub Enterprise Server, making it suitable for teams of all sizes. GitHub
Sourcebot MCP: Sourcebot MCP server enables LLM agents to fetch code context from various repositories, enhancing responses to user requests and improving reasoning for existing agents. This is particularly usefuul, to get a better context for your LLMs, to give specific code recommendations. Sourcebot MCP
Filesystem MCP Server offers secure file operations with configurable access controls, serving as the foundation for most development workflows. It provides sandboxed directory access with path validation and traversal protection, ensuring security while enabling powerful file manipulation capabilities. Filesystem MCP
Database integration servers bridges the gap between natural language queries and data operations. The PostgreSQL MCP Server enables read-only database queries with intelligent schema inspection, Cursor GitHub while the MongoDB MCP Server provides document database operations with Atlas API integration. you will find a host of database integration servers at awesome-mcp-server
Exa Search: Exa Search enhances AI models’ ability to retrieve accurate and relevant information from the web in real-time. It performs recursive searches to gather comprehensive data, provides structured search results, including titles, URLs, and content snippets.Exa Search
Cloud platform integrations
AWS MCP Servers represent the most comprehensive cloud integration, with specialized servers for different services including Bedrock AI, CDK infrastructure management, Lambda functions, and DynamoDB operations. These servers enable developers to manage complex cloud infrastructure through conversational interfaces.
Azure MCP Server provides key Azure services integration including Cosmos DB, Storage, and Log Analytics, with proper RBAC permission handling for enterprise security requirements. GitHub
K8s MCP Server: Kubernetes Model Context Protocol (MCP) server provides tools for interacting with Kubernetes clusters. It offers API resource discovery, resource listing, detailed information, descriptions, pod logs, node and pod metrics, event listing, and resource creation/updating. K8s
Testing, Web scraping
Semgrep MCP: MCP server provides access to Semgrep’s static analysis, including security checks, secret detection, code quality checks, and custom rules. Semgrep’s engine is transparent, extensible, and fast, making it easy for LLMs to interface with Semgrep .SemGrepMCP.
Puppeteer: This MCP server provides automation for controlling your browser, very useful for performing web testing. last time I checked, it is being archived!!! not sure why. But still a good tool to check, this is very close to the next one in the list.Puppeteer
Browserbase MCP: Browserbase MCP allows AI models to interact with live web pages, enabling tasks like data extraction, DOM inspection, and user action simulation. This MCP server can execute JavaScript within a cloud browser environment, capture screenshots and interacts with web elements and facilitate real-time web debugging and automation. Browserbase
Firecrawl: Firecrawl MCP provides robust web scraping capabilities, delivering LLM-ready data directly into your development environment. It supports complex scraping rules and batch processing, delivers data in formats like Markdown or JSON and as the website claims, it can process up to 100 pages per minute. This can be used to extract information from the websites. The most powerful is the search tool(firecrawl_search), which can be used to search for specific information. Firecrawl
Slack MCP Server transforms team communication by enabling automated messaging, channel management, and notification systems directly through AI interactions. This creates seamless integration between development workflows and team collaboration.
Design
UML MCP: UML-MCP is a diagram generation server that supports various diagram types, integrates with MCP-enabled LLM assistants, and offers multiple output formats. UMLMCP
Magic MCP: This provides a AI-powered UI generation, enabling you to create UI components simply by describing them in natural language. You’ll have access to a modern component library, featuring a vast collection of pre-built, customizable components inspired by 21st.dev. With real-time preview capabilities, you can see your components as you create them. It also provides full TypeScript support for type-safe development and integrates with SVGL, giving you access to a comprehensive collection of professional brand assets and logos. MagicMCP
Figma MCP integration enables remarkable productivity gains. This server has tools to directly manipulate design files, extract assets, and generate corresponding code. This is by far the best set of tools, that helped me accelerate my design to code time. FigmaMCP
Blender 3D MCP enables complex 3D scene creation through natural language. THis is one of the crazy servers, I tried :-) it was so much fun to generate 3D models, though I am not into gaming or AR/MR kind of app development. But this is going to be very powerful for future AR/MR/VR kind of applications. BlenderMCP
Its raining MCP servers…
There are host of mcp servers on web. Some are official like Official Model Context Protocol, Cursor Official Registry, some are community built and managed. awesome mcp servers site is a very good community of mcp servers.
I also strong recommend checking out Zapier, which provides a good hosted option, and single interface to a host of MCP servers.
Just remember to use what is rated well, and used/supported by larger community. Make sure these are safe. Remember some of these MCP servers request for critical information like API keys, access to your local repository ot GIT repositories.
Each of these tools brought me closer to what I think of as true vibe coding — fast feedback loops, less friction, and more room for creative expression. With the right MCP stack, I can go from idea → plan → code → test → refine in a single sitting, without killing momentum.
There you go…there are host of other servers, that I have been exploring and integrating into my workflows. I will soon blog about the end SDLC, not just coding…
Leave your experiences, and what are the top MCP servers you are using, to help the readers…:-D
