Introduction : From Learning to Doing
If Part 1 of this series was the theory understanding LLMs, function calling, and the magic of MCP then Part 2 is the field report.
Recently, I attended the NxtWave MegaWorkshop, a mind-expanding deep dive into MCP and its possibilities. It wasn’t just a “sit and listen” event it was an I left with more tabs open in my brain than in my browser kind of workshop.
I walked in curious. I walked out with:
- A stack of insights that changed the way I think about AI.
- A toolbox full of MCP-powered skills.
- And a list of projects I never thought I’d build in days instead of months.
So, I hope you’ve read Part 1 because this time, it’s all about experience and implementation. This blog might run a little long (worth every scroll) because I’ll be walking you through:
- My deep dive into Cursor and how it supercharged my workflow.
- How I integrated Pipedream for powerful automation.
- The workshop tasks and projects I completed.
- And, out of pure curiosity, how I went on to build my full-stack project “TaskAura” entirely using Cursor the challenges, the wins, and the problem-solving tricks I picked up.
Think of this as both a story and a playbook the things I built, the problems I ran into, and the tips I wish I’d known earlier.
The Workshop Experience : Where Theory Met Reality
In this workshop, I learned a lot of things but most of the theory I’ve already covered in Part 1. Now it’s time to talk about the real-world projects and tasks I completed.
I built 5 different projects:
- Project 1: LinkedIn Post Creation with Image
- Project 2: Gmail Automation
- Project 3: Railway Train Schedule
- Project 4: Eleven Labs Voice Generation (with Twilio)
- Project 5: Learning Path Generator
You might be thinking these don’t sound that impressive. But here’s the thing: sometimes it’s these small, repetitive tasks that eat away at your time and patience. MCP turns them into one-click (or even zero-click) automations. These are the basic use cases that quietly remove friction from your day, letting you focus on work that actually matters.
The NxtWave MegaWorkshop wasn’t your usual “sit in a chair, take notes, go home” kind of event. This was hands-on chaos meets guided brilliance. One moment we were learning how MCP could turn an AI into a real-world operator, the next moment we were building live integrations that actually… worked.
From the very first session, the emphasis was clear no passive learning. Every concept came with a challenge, every challenge with a build. You weren’t just told how MCP works you had to prove it by making it dance.
Highlights from the sessions:
- Watching function calls light up real APIs in real time.
- Deploying quick automations that made me rethink my “manual” workflows.
- Connecting tools I’d used for years in ways I didn’t think were possible.
By the end of the workshop, my brain felt like a Git repository that had just been force-pushed with a thousand new commits.
And this wasn’t just about MCP in isolation the facilitators introduced Cursor for AI-assisted coding and Pipedream for lightning-fast integrations, both of which became my MVPs for the projects that followed.
Tool #1 — Cursor: My AI Pair Programmer on Steroids
If MCP is the engine, Cursor is the high-performance steering wheel I didn’t know I needed. In MCP terms, Cursor acts as the MCP client it connects to one or more MCP servers to access tools, resources, or capabilities. Your desktop machine becomes the host running Cursor, while the servers supply the functions it can trigger.
In my personal opinion, Cursor is one of the easiest AI coding tools to pick up and run with and yet it’s powerful enough to feel like you’ve just hired a full-time developer who never sleeps.
Why Cursor stood out for me:
- Ridiculously easy to use: the interface feels familiar if you’ve used VS Code, but smarter out of the box.
- Seamless integration with MCP: adding new tools or APIs feels less like “configuration” and more like “click and go.”
- Real-time AI assistance: not just suggesting code, but understanding context, fixing bugs, and improving logic.
- Rapid prototyping: I could build, test, and refine projects in a fraction of the time compared to my usual workflow.
I built all five projects from the workshop using only the free trial credits. That’s how efficient it was even without spending a dime, I got a lot done.
The one downside? It’s a paid tool once the free credits are over. I stopped after finishing my free trial for now, but I’m already planning my next round of builds once I subscribe.
And yes, while Cursor is my current favorite, there are other interesting players in the space like Augment and Zencoder which are free and on my list to explore next.
Tool #2 — Pipedream
Think of Pipedream as Zapier on steroids with a developer soul it’s a workflow automation tool that’s infinitely flexible.
Where Zapier gives you no-code, Pipedream gives you low-code + full code freedom, meaning you can stitch APIs together, run custom Node.js/Python scripts, and orchestrate entire AI workflows in minutes.
Why it’s innovation-friendly
- Rapid Prototyping: Launch MVP automations without setting up servers or deployment pipelines.
- Hybrid Workflows: Drag-drop integrations plus inject custom code wherever needed.
- Event-Driven: Trigger workflows from webhooks, timers, databases, or even Slack commands.
- Native AI Integrations: OpenAI, Hugging Face, LangChain, Pinecone, and more all plug-and-play.
- Cost Efficiency: Pay for usage, not bloated SaaS overhead.
How you could use it in your innovation playbook
- AI-powered market scanning Schedule Pipedream to scrape niche news + run through GPT for insights → push summary to Slack every morning.
- Real-time customer feedback triage Hook into Google Forms/Typeform → sentiment analysis with OpenAI → auto-route to the right team.
- Prototype data-driven products Ingest data from APIs or Google Sheets → transform via code step → push to dashboards or ML models.
- Connect siloed tools instantly Sync Notion, Airtable, HubSpot, and GitHub in one event-driven flow no messy custom backend needed.
If Make (Tool #1) is your visual LEGO builder, then Pipedream is your Swiss Army knife for high-velocity, code-augmented automations.
Toolchain Setup : Connect Cursor + Pipedream (Quick Start)
- Install & sign in to Cursor Download, open, and log in (Google/GitHub). That’s it.
- Get an MCP server from Pipedream Go to Pipedream → sign in → search a tool (e.g., Gmail/LinkedIn) → Connect Account → open the Cursor tab → copy the MCP config shown there (your authenticated URL + JSON).
- Add the server to Cursor Cursor → Settings → Tools & Integrations → Add Custom MCP → paste the config → Ctrl+S (save).
-
Use it from the AI panel
Open New Chat (or Ctrl + Alt + B) and talk to your tools, e.g.:
- “Send an email with the subject ‘Status Update’ to my team.”
- “Post this text + image to LinkedIn.”
- “Get Hyderabad → Tirupati trains on 2025-07-21.”
- Rinse & repeat for more tools Add more MCP servers (YouTube Data, Google Drive, Notion, Railway, etc.) by repeating steps 2–3.
Security reminder
- Do not share your Pipedream MCP URLs. They’re auth’d to your accounts.
If Pipedream hiccups
- Try the Composio MCP server as a drop-in alternative.
In MCP terms: Cursor acts as the MCP client; Pipedream provides the MCP servers; your machine is the host. Plug in a server, and your AI gains a new superpower.
🚀 My MCP Workshop Projects
As promised, here’s the real-world action from Part 1’s theory.
These are the projects I built during the workshop using Cursor + Pipedream.
I can’t share all the screenshots here (or this blog would become a photo album), but trust me it’s simple enough for you to replicate. I’ll drop a few reference shots, and you can take it from there.
Project 1: LinkedIn Post Creation with Image
The AI drafts the content, generates an image, and posts it straight to LinkedIn.
No logging in, no copy-pasting just:
Linkedin Implementation
Create an image post on LinkedIn with the following text:
{Your post text}
Image link: {Image link}
And boom 💥 it’s live.
Here’s the actual post I made using this: 🔗 [My LinkedIn Post]
Project 2: Gmail Automation
From sending polished emails to fetching my latest inbox updates all triggered by plain English.
Example:
“Get my details from LinkedIn, apply for a job at {company name}, write a short cover letter, and send everything to {mail}.”
Project 3: Railway Train Schedule
Ask for train details and MCP instantly returns routes, times, and availability.
Prompt:
get the train info: KRISHNA EXP
Prompt:
Get seat availability of KRISHNA EXP
Prompt:
Get train live status on date: 2025-07-17
train number: 17230 (SABARI EXPRESS)
Project 4: Eleven Labs Voice Generation (with Twilio)
This one’s like giving your AI a literal voice.
I connected Eleven Labs (for lifelike AI speech) with Twilio (for making calls). The setup involved:
- Creating an API key in Eleven Labs.
- Buying a free-trial Twilio number.
- Linking both so MCP could send voice updates.
The result? I built a “Tech Update Agent” that could call my phone and deliver the latest AI, programming, and cybersecurity news in a natural, friendly voice.
Think of it as an AI friend ringing you up with news instead of sending a newsletter.
Prompt 1:
Create an agent that makes an outbound call to update someone about recent tech news
Use a confident, friendly tone — like a helpful colleague
Deliver updates about the latest in AI, programming, and cybersecurity
Keep the explanation short, clear, and jargon-free
Use my voice (female) for the call
The agent should sound like a tech-savvy friend, not a sales rep
First message: "Hey, I’ve got some quick tech updates for you — should I go ahead?"
You don’t have to make a call. Create an agent in the ElevenLabs simply
The agent will be created in Eleven Labs.
Prompt 2:
Now make an outbound call to my number, ask for my number
Project 5: Learning Path Generator
Type your goal (e.g., “Learn full-stack web development in 6 months”), and MCP maps out a tailored step-by-step learning plan.
Using these workflows, I even posted live to LinkedIn straight from my AI—no tab-switching, no hassle. The real magic? It’s not that each task is impossible manually… it’s that doing them without touching the browser feels like stepping into the future.
Taskaura : The Real Monster Out of the Cage
So, let’s bring the real monster out of the cage.
Out of pure curiosity, I ended up building a full-stack app “Taskaura” using ReactJS, Tailwind, Node.js, and MongoDB. And here’s the kicker: I did it almost entirely through vibe coding with MCP. No deep dives into code, no manual debugging marathons just high-quality prompts, a bit of persistence, and three hours later… I had something real.
The Idea
For months, I’d been wanting a personal tracking app for my daily routines and learning habits. My personal motto:
“Learn at least one new thing every day and document it.”
The problem? Training commitments left me no time to actually build it. But when MCP landed in my hands, I thought: Why not let it do the heavy lifting?
The App
Taskaura has four main pages:
- Dashboard — Shows my streak, learning rate, and overall progress at a glance.
- Daily Page — Like a to-do list. Each completed item increases my score.
- Weekly Page — Works like Daily, but resets every week.
- Learning Page — A log of everything new I’ve learned, plus my learning rate and streak to keep me motivated.
The Experience
Honestly? Mind-blowing.
From creating the app to deploying it, MCP handled it all. I just fed it prompts and boom, code happened. But here’s the truth:
- The core functionality came together in about 3 hours.
- The real work was fixing UI bugs, backend integration headaches, and styling quirks… which took me another week.
The Mistakes
I learned the hard way:
-
Changing stack mid-project was chaos.
- Started in React (frontend only), then decided to add a backend for login and data storage.
- Connecting backend to frontend ate up a lot of time.
- Thought of migrating to Next.js (to avoid separate backend)… huge mess. Lost all styles, broke layouts, and had to revert from Git.
-
Prompts that were too big.
- Gave huge instructions in one go. MCP choked. Iterative prompting would have been faster and cleaner.
You're working in a Vite + React + TypeScript + Tailwind CSS project with Framer Motion and Chart.js installed.
🚀 Build a productivity dashboard named **TaskAura** with the following features:
---
### 1. 🌐 Route: `/dashboard` (Main Page)
**Purpose:** Show summary of all productivity features.
Build a `Dashboard.tsx` page with:
- A radial or bar chart showing % of weekly tasks completed
- Section showing count of daily tasks done vs total (today only)
- Section showing current learning streak (🔥 emoji + count)
- Latest 3 learnings from streak with title and category badge
- A daily motivational quote at the top
- All cards should use Framer Motion (`slideIn`, `fadeIn`)
- Responsive grid layout: 1-column mobile, 3-columns desktop
- Use Tailwind's `dark:bg-dark-bg` and `text-light-text` classes
---
### 2. 🗂 Route: `/weekly` (Weekly Task Planner)
**Purpose:** Add/edit/delete/track weekly tasks.
Build a `Weekly.tsx` page that:
- Displays weekly tasks from localStorage key `weeklyTasks`
- Each task should have:
- Title, optional description, checkbox, and delete icon
- Progress bar or chart visualizing completion
- Button to add new task using modal with:
- Title (required)
- Description (optional)
- "Reset Week" button that clears all weekly tasks
- Confetti animation when all tasks are marked done
- Animate task check with Framer Motion
---
### 3. 📅 Route: `/daily` (Daily Tasks)
**Purpose:** Track today-only tasks, lightweight view.
Build a `Daily.tsx` page that:
- Loads daily tasks from localStorage key `dailyTasks-{YYYY-MM-DD}`
- Displays a list of tasks (title + checkbox + delete)
- Floating action button (FAB) opens modal to add a task
- Modal fields: Title (required), Note (optional)
- Show daily progress bar (% done)
- Optional: Inspirational quote at the top
- All tasks should animate in when added or marked
---
### 4. 📚 Route: `/learn` (One Day One Learn)
**Purpose:** Track what I learn every day and build streaks.
Build a `Learn.tsx` page that:
- Contains a form with the following fields:
- `What you learned`: text input (required)
- `Description`: textarea
- `Category`: dropdown (Tech, Life, Finance, Mindset, Other)
- `Source`: text or URL (optional)
- Only allow one submission per day (`learnHistory` in localStorage)
- Show 🔥 streak bar:
- Count of consecutive days
- Reset to 0 if no entry added today
- Below form, render a timeline (most recent learnings at top):
- Card: Title, category badge, source link, date
- Confetti/fire animation on streak milestones (e.g., 7, 30)
---
### 5. 🌘 Dark/Light Mode Toggle
- Build a persistent theme toggle (`ThemeToggle.tsx`)
- Use Tailwind’s `dark` class strategy (`class="dark"`)
- Save theme preference in `localStorage`
- Animate toggle switch on click
---
### 6. 📊 Reusable Components
Build reusable components:
- `ChartCard.tsx` → Takes `title`, `data`, `type`, renders bar/doughnut
- `AddTaskModal.tsx` → Reusable for weekly/daily tasks
- `StreakDisplay.tsx` → Shows 🔥 icon + current streak count
- `LearnTimeline.tsx` → Displays list of past learnings
- `MotivationalQuote.tsx` → Returns random quote from local list
---
### 7. 🧠 Storage Helpers
In `storage.ts`:
- `saveToStorage(key: string, value: any): void`
- `loadFromStorage(key: string): any[]`
- `getTodayKey(): string` → returns e.g. `dailyTasks-2025-07-25`
- `updateStreak()` → calculate and return new streak
---
📦 All data must persist using **`localStorage`**, no backend.
⚡ Use Tailwind CSS for styling, Framer Motion for animations, and Chart.js for charts.
🎯 The result should be a beautiful, responsive, animated productivity app that helps plan weekly goals, focus daily, and build a learning habit called **TaskAura**.
This was the initial prompt I gave, and I iteratively made it work. But instead of doing it this way, it’s better to break the task into parts and specify each in depth. The app was initially implemented using localStorage, and later enhanced with backend integration.
-
Blind trust in the AI’s code.
- Some code was too advanced for me to instantly grasp. When it didn’t work, debugging took forever.
The Challenges
- Backend integration delays.
- AI not always following instructions exactly.
- UI fixes eating up more time than expected.
Suggestions (from the trenches)
- Pick your tech stack from the start. No mid-project switches unless absolutely necessary.
- Use iterative prompting—break tasks into steps, not essays.
- Check DevTools early and often; fix issues progressively.
- Keep your repo clean—commit often so you can revert to working states.
- Don’t be afraid to ask AI to explain its own code—understanding beats copy-pasting.
- When possible, design before you code—even a rough sketch saves hours.
Why This Was Different
Taskaura wasn’t just another project. It was the first time I felt like I could go from idea → MVP → live app in the span of a weekend and it’s all because MCP turned coding from a skill problem into a prompting problem.
Wrapping Up: MCP — Not Magic, Just Smarter Work
When I began this journey, I thought MCP was just another tech buzzword. Now? It’s my AI’s passport to the real world thanks to NxtWave for opening this gateway. Their MegaWorkshop didn’t just teach me MCP; it reshaped how I approach automation.
From LinkedIn posts to Taskaura, MCP transformed my AI from a chatty theorist into a doer. No more “I can’t access that” just “Tell me what to do, and consider it done.”
Key Takeaways
- Start small, chain big: Begin with one tool (like Gmail), then scale.
- Cursor + Pipedream = Superpowers: The ultimate no-code duo for rapid automation.
- Mistakes are fuel: My backend chaos taught me to plan stacks upfront.
- Prompt iteratively: Baby steps > monolithic prompts.
MCP isn’t about replacing effort it’s about redirecting it. Less time wrestling APIs, more time building what matters.
Part 1 was the “what.” Part 2 was the “how.” Now? It’s your turn. Pick a tool, plug it into MCP, and watch your AI earn its keep.
A huge shoutout to NxtWave for introducing me to this game-changer.
P.S. Connect with me on [LinkedIn] I’d love to see what you automate first! 🚀
🔗 Connect with Me
📖 Blog by Naresh B. A.
👨💻 Aspiring Full Stack Developer | Passionate about Machine Learning and AI Innovation
🌐 Portfolio: [Naresh B A]
📫 Let's connect on [LinkedIn] | GitHub: [Naresh B A]
💡 Thanks for reading! If you found this helpful, drop a like or share a comment feedback keeps the learning alive.
Top comments (0)