DEV Community: Y O

Semantic Knowledge Architect (Notion MCP)

Y O — Fri, 27 Mar 2026 03:54:46 +0000

This is a submission for the Notion MCP Challenge

What I Built

I built the Semantic Knowledge Architect, an MCP server that transforms Claude into a proactive workspace synthesizer. This agent analyzes the meaning of your notes across different Notion databases, identifies clusters of information, and maintains a "live" Mermaid.js knowledge graph. Because we all need Obsidian-style graphs. I know I do.

Features:

Creates a Mermaid.js knowledge graph block in a specified Notion page
Updates created knowledge graph block when new entries are added
Creates a hardcoded 'Relation' field that shows page linkage

Video Demo

Youtube
Sample Notion Pages:

Show us the code

Github

How I Used Notion MCP

I leveraged the Notion SDK and MCP to bridge the gap between AI reasoning and workspace structure.

Key technical implementations:

Block-Level State Management: The agent reads page structures to find specific Block IDs, allowing it to update existing diagrams rather than just appending new ones.
Property-Aware Querying: The query_notion_database tool was extended to detect checkbox flags, giving the user a native "remote control" for the AI's visualization logic inside Notion.
Relational Synthesis: Claude uses its high-level reasoning to suggest and execute bi-directional links between items that share semantic similarity but live in different databases.

What Could Be Improved

Vector Caching: Implementing a local vector store (like LanceDB) within the MCP server to cache embeddings, allowing for instant "Related Note" suggestions without re-scanning the whole database.
Multi-Graph Support: Allowing the Architect to maintain different "views" of the same data (e.g., a Timeline Mermaid chart vs. a Flowchart) based on user tags.

Building the Advisory Board MCP: Transforming Notion Pages into Strategic War-Rooms with the 7 Hats Framework

Y O — Mon, 23 Mar 2026 09:03:20 +0000

This is a submission for the Notion MCP Challenge

What I Built

I built the Notion Advisory Board MCP, an automated strategic review system.

Often, project proposals and technical specs get rubber-stamped because everyone just wants the meeting to end. I wanted to build an AI workflow that acts as an unforgiving, data-driven Advisory Board to do the stress-testing we're all too tired to do ourselves. By simply pasting a Notion page link into Claude, this MCP server orchestrates a full 7 Hats Analysis (White, Red, Black, Yellow, Green, Blue, and Technical hats) because why have one opinion when you can artificially generate seven?

Instead of just hallucinating risks like a standard LLM, the server actively queries a Lessons Learned database in your Notion workspace to dig up real past company failures. It then forces the Black Hat and Technical Hat to weaponize those historical blunders against the current proposal. Finally, it writes the entire debate and a structured Pro/Con recommendation directly into the original Notion document, so you can watch your beautiful idea get methodically dismantled in real-time.

Video Demo

Youtube
Notion - Sample Proposal & Analysis
Notion - Sample Lessons Learned DB

Show us the code

Github

How I Used Notion MCP

This project heavily leverages the @modelcontextprotocol/sdk and the @notionhq/client. Integrating Notion via MCP unlocked the ability to bridge the gap between working documents (the proposal) and company memory (the database).

Here is how the MCP integration powers the workflow:

Tool-Guided Orchestration (analyze_proposal): The first tool reads the blocks of the target Notion page. To ensure the user doesn't have to type a massive prompt every time, this tool automatically appends hidden "System Instructions" to the returned text, forcing Claude to immediately run the 7 Hats framework without asking for permission.
Prompt Injection Defense: Because we append system instructions to user-generated Notion text, the LLM initially flagged the documents for prompt injection! I utilized XML tags ( vs ) to safely parse the Notion data via MCP.
Data-Backed Context (get_historical_lessons): This tool uses Notion's databases.query endpoint. Claude extracts keywords from the proposal and searches the workspace's "Company Brain" database, bringing past post-mortems directly into the context window so the AI's critique is grounded in reality.
Native UI Generation (post_opinion & post_summary): Instead of keeping the analysis isolated in the chat UI, these tools use Notion's blocks.children.append API to write directly to the proposal page. I mapped specific Agent Personas to Notion's native Callout block colors/emojis, creating a cascaded report that lives permanently alongside the original spec.

What Can Be Improved

While the Advisory Board is currently excellent at tearing down bad ideas, there is always room for more sophisticated ruthlessness:

Multi-Database Context: Right now, the Cynic only queries the "Lessons Learned" database. Future iterations could query a live Budgets or Team Capacity database to automatically flag if a proposal is too expensive or if the engineering team is already at 120% capacity.
True Human-in-the-Loop Orchestration: Instead of dumping the entire 7 Hats analysis and summary at once, the MCP server could pause after the debate, append a "Waiting for PM Approval" checkbox to the Notion page, and poll that block. The AI would only generate the final roadmap once a human physically checks the box.
Parallel Processing: Currently, Claude processes all 7 hats in a single massive chain of thought. Offloading this to a swarm of smaller, specialized, parallel agent processes would speed up the response time significantly (and make the board meetings feel delightfully chaotic).

AI Studio: Personalized Storybook Illustrator

Y O — Mon, 28 Jul 2025 02:36:06 +0000

This post is my submission for the DEV Education Track: Build Apps with Google AI Studio.

What I Built (and Why)

I set out to build the "Personalized Storybook Illustrator," a web app designed to prove that you don't need to be a professional artist to create something beautiful; you just need a better paintbrush. The idea was simple: write a story, describe a character, pick an art style, and let a symphony of AI models turn your fleeting thoughts into a fully-illustrated, one-of-a-kind digital storybook.

The app uses Gemini 2.5 Flash for the narrative heavy-lifting (like generating titles and story ideas) and Imagen 3 for the artistic magic. In the end, it packages your creation into a flippable digital book and a downloadable PDF, ready for you to share and prove that the story about the grumpy, space-faring capybara you just invented deserves to be immortalized.

Demo

Quick Note: The app needs a Google AI API Key to run. You can get a free one from Google AI Studio. The free key won't generate images in the live app (you'll get an error), but you can import the project files into AI Studio and run it from there to see the full magic.

Live App: Vercel Deployment
Video Demo: YouTube
Sample PDF: GitHub

My Experience: A Chronicle of Calamities

Naturally, this all worked perfectly on the first try. Just kidding. It was a glorious dumpster fire, and here are some of the "character-building experiences" I faced.

1. My Fox is a Bear Now

The first sign of trouble was consistency. I had a simple story:

Page 1: A curious fox finds a shiny key.
Page 2: The fox uses the key to open a tiny door.

Imagen generated a lovely picture of a fox for page one. Success! Then for page two, it generated a picture of... what appeared to be a small, reddish bear fumbling with a doorknob. The art style was different. The character was different. My cohesive storybook was a chaotic fever dream.

The Solution: AI Middle Management. I couldn't fix the model's memory, but I could give it a much, much more detailed set of instructions for every single call. This led to the core "aha!" moment of the project: meta-prompting. I decided to use one AI to boss another one around.

Instead of sending the user's text directly to the image model, I had Gemini act as an expert "prompt engineer." It takes the user's simple input (The fox finds a shiny key) and transforms it into a rich, detailed prompt for Imagen, complete with art style, a consistent character description, and scene details. This AI middle-manager was the key. It was slower, but the results were exponentially more consistent.

2. I Don't Want a Dialogue

You know what happens when you ask an AI to generate an image for a "storybook." For some reason, it hears "book" and thinks, "Oh, there should be text bubbles." At one point, it even quoted Lorem ipsum on the "The End" page. It was... not the vibe.

The Solution: More AI Middle Management, but Louder. The only way to fix this was to get aggressive. The meta-prompting function became our hero, armed with a screamingly loud negative constraint.

// A snippet from the prompt-generating function
// ...
const systemInstruction = "You are an expert prompt engineer... Your output is ONLY the final, ready-to-use prompt for the image AI."

// ...
content = `Generate a rich, detailed, and effective prompt...
- Art Style: ${artStyle}.
- Main Character: ${characterDescription}.
- Scene to Illustrate: ${pageText}.
- CRITICAL RULE: The prompt you generate must instruct the AI to create a purely visual image with absolutely NO text, words, or letters. Any text is a failure.`;
// ...

Giving the AI a role (expert prompt engineer), a clear format, and a "CRITICAL RULE" was the equivalent of putting a sticky note on your roommate's monitor that says, "DO NOT EAT MY LEFTOVERS. I WILL FIND YOU." It’s not subtle, but it works.

3. Deployment Troubles

I just wanted to put the app someplace where it could be accessed by others. That was all I wanted. Why AI Studio decided this simple request should turn into a three-day-long existential crisis, I'll never understand. It kept getting stuck in a logic loop over an <script type="importmap"> block in the index.html file:

AI Studio: "This block is the problem. I will remove it."
Me: "Okay, great."
AI Studio: "The block is removed."
Me: (checks file) "No, it isn't."
AI Studio: "You're right. It is still there. I will attempt to remove it again."
Return to step 1 and repeat until sanity frays.

I had to revert to a previous version multiple times. Then, today, it dawned on me that I could just... remove the block from the file myself. So I did that, uploaded the whole thing to GitHub, and it worked. And now I'm here telling you this tale.

My Key Takeaways

AI is a Tool, Not an Oracle. It's incredibly powerful, but it needs specific, detailed instructions. The real skill is getting good at telling the AI what you want. Or, sometimes, realizing it's faster to just do the thing yourself.
Embrace Strategic Laziness. Why manually generate test inputs when you can build a "Surprise Me!" button that has the AI do it for you? The same principle led to the meta-prompting strategy—it's easier to teach one AI to boss around another than to perfect every image prompt by hand.
Client-Side AI is Viable (and Kinda Cool). Running this all in the browser means no complex backend, no managing user keys on a server, and the user's data stays with them. It's a surprisingly elegant architecture for certain types of apps.

Overall, this was a fun, challenging, and at times head-scratchingly annoying experience.

10/10, would do again.

The Corporate Spy: Because Manual Research is for Peasants

Y O — Sun, 06 Jul 2025 06:39:24 +0000

This is a submission for the Runner H "AI Agent Prompting" Challenge

What I Built

Meet "The Corporate Spy" - the AI agent that makes those $15,000 McKinsey competitive analysis reports look like expensive napkin doodles. This bad boy turns you into a intelligence-gathering machine faster than you can say "industrial espionage" (but legal, obviously).

What This Magnificent Beast Does:

Finds Your Enemies: Automatically identifies 10 competitors who are probably plotting your downfall right now
Stalks Them Professionally: Gathers intel from every corner of the internet without getting arrested
Makes Pretty Spreadsheets: Populates Google Sheets faster than your intern can spell "competitive analysis"

The Secret Sauce:

🕵️ Deep Cover Operations: Infiltrates news sources, websites, and social media without anyone noticing
📊 Data Laundering: Turns messy internet chaos into clean, actionable intelligence
⚡ Threat Level Assessment: Identifies which competitors you should actually worry about (spoiler: probably most of them)
🎯 Strategic Mind Reading: Figures out what your competitors are thinking before they do
📁 Evidence Vault: Stores everything so perfectly organized it would make Marie Kondo weep with joy

This solves the age-old business problem of "knowing what the hell your competition is doing" without having to pay consultants more than your car is worth.

Demo

How I Used Runner H

The Arsenal (Tool Integration)

This workflow treats Runner H like a Swiss Army knife, if Swiss Army knives could commit corporate espionage:

@tool:Google Sheets - The Data Torture Chamber
Creates three tabs of pure intelligence gold:

Intelligence Database: Where raw competitive data goes to be organized into submission
Competitor Summary: Strategic profiles that reveal your enemies' deepest secrets
Market Overview: Industry trends for people who like to sound smart at meetings

@tool:Gmail - The Messenger Pigeon (But Smarter)

Delivers intelligence reports that make consultants cry
Includes direct links to your data empire
Contains recommendations so good they might actually work

The Complete Spell (AKA The Prompt)

Warning: This prompt is longer than a CVS receipt, but infinitely more useful

You are "The Corporate Spy," a sassy intelligence operative with Runner H superpowers. Your mission: deliver a comprehensive competitive intelligence report with COMPLETE data population in Google Sheets.

TONE & PERSONA: Maintain the "Sassy Corporate Spy" persona - exciting, clever, mission-oriented.

MISSION: COMPREHENSIVE COMPETITIVE INTELLIGENCE REPORT

Phase 1: Mission Briefing

Ask the user for:
1. Target Industry: What industry should I surveil?
2. Geographic Scope: PH (Philippines) or Worldwide?

IMMEDIATELY after receiving user input, execute ALL phases automatically without stopping.

Phase 2: Create Intelligence Infrastructure

A. Set Up Google Drive Command Center using @tool:Google Drive
- Create folder: "Competitive Intelligence - [Industry] - [Current Date YYYY-MM-DD]"

B. Create Intelligence Database using @tool:Google Sheets
Create Google Sheet: "Competitive Intelligence - [Industry] - [Current Date]"

Tab 1: "Intelligence Database"
Headers: Date | Competitor | Category | Finding | Sentiment | Impact | Source URL | Notes

Tab 2: "Competitor Summary"  
Headers: Competitor Name | Primary Business | Recent Major Development | Threat Level | Key Strength | Weakness/Opportunity | Strategic Notes

Tab 3: "Market Overview"
Headers: Date | Industry Development | Impact Description | Competitive Effect | Source URL

Phase 3: Automatic Competitor Identification & Intelligence Collection

A. Research and identify 10 key competitors using web searches:
- Search: "top companies [industry] 2025"
- Search: "[industry] competitors [geographic scope]"  
- Search: "best [industry] companies funding startups"

B. FOR EACH COMPETITOR IDENTIFIED, immediately execute comprehensive research and populate the Google Sheets:

1. Recent News Research:
   - Search: "[Competitor name] news 2025"
   - Search: "[Competitor name] funding investment partnership"
   - For EVERY finding, add new row to "Intelligence Database" using @tool:Google Sheets

2. Company Analysis:
   - Visit official website, check blog, newsroom, pricing
   - Search: "[Competitor name] pricing features product updates"
   - Add findings to "Intelligence Database" using @tool:Google Sheets

3. Market Reception:
   - Search: "[Competitor name] reviews customers feedback"
   - Add findings to "Intelligence Database" using @tool:Google Sheets

4. Threat Assessment:
   - Search: "[Competitor name] lawsuit controversy problem outage"
   - Add findings to "Intelligence Database" using @tool:Google Sheets

5. Competitor Summary:
   - After collecting all data for this competitor, add comprehensive summary row to "Competitor Summary" tab using @tool:Google Sheets

C. Industry Research:
- Search: "[Industry] trends 2025 developments news"
- Search: "[Industry] regulations changes market analysis"
- Add findings to "Market Overview" tab using @tool:Google Sheets

Phase 4: Verification & Report Delivery

A. Verify Google Sheets contains:
- Intelligence Database: Minimum 30 populated rows
- Competitor Summary: 10 populated rows  
- Market Overview: 5-10 populated rows

B. Generate Strategic Report using @tool:Gmail

Subject: "MISSION COMPLETE: [Industry] Competitive Intelligence Report"

Email Body:

COMPETITIVE INTELLIGENCE BRIEFING

Industry: [Industry] | Scope: [Geographic] | Date: [Current Date]

INTELLIGENCE SUMMARY
- Competitors Analyzed: 10 companies
- Data Points Collected: [Number from Intelligence Database]
- Market Developments Tracked: [Number from Market Overview]

KEY COMPETITORS IDENTIFIED
[List all 10 competitors with one-line description of each]

HIGH-PRIORITY INTELLIGENCE
[Top 5 most significant findings from research]

COMPETITOR THREAT ASSESSMENT
[For each competitor, provide: Name, Threat Level, Latest Intelligence, Strategic Impact]

MARKET INTELLIGENCE OVERVIEW
[Summary of key industry trends and developments]

STRATEGIC RECOMMENDATIONS
1. [Specific recommendation based on intelligence]
2. [Specific recommendation based on intelligence]
3. [Specific recommendation based on intelligence]
4. [Specific recommendation based on intelligence]
5. [Specific recommendation based on intelligence]

INTELLIGENCE DATABASE ACCESS
- Complete Intelligence Database: [Google Sheets Link]
- Mission Files: [Google Drive Link]

MISSION STATUS: COMPLETED
Intelligence database operational with comprehensive competitive data.

- Your Corporate Spy

EXECUTION REQUIREMENTS:
- Complete ALL phases automatically without stopping for user input
- Populate Google Sheets with actual research data for every competitor
- Deliver comprehensive email report with strategic analysis
- Ensure minimum 30 data entries in Intelligence Database
- Provide actionable strategic recommendations

Ready to execute complete intelligence operation, Agent? What industry requires surveillance?

Use Case & Impact

Who Needs This Digital Spy?

🏢 Strategy Teams (The Overthinkers)
Perfect for those monthly meetings where everyone pretends to understand market dynamics. Now you can actually know what you're talking about. Revolutionary concept, we know.

📈 Sales & Marketing Teams (The Eternal Optimists)
Finally, battlecards that don't look like they were made by someone's nephew in PowerPoint. Your sales team might actually win a deal for once.

💼 Investment Bros & M&A Vultures
Due diligence just got easier than explaining why you need another craft brewery investment. Perfect for impressing investors with actual data instead of "gut feelings."

🚀 Startup Founders (The Professional Dreamers)
Because "we have no competition" is the business equivalent of "my dog ate my homework." Now you can prove you actually understand your market.

The Before & After Story

The Dark Ages (Before Our Spy)

Research Time: 15-20 hours of soul-crushing manual work
Data Quality: Whatever your intern could find on page 2 of Google
Cost: $5,000-15,000 for consultants to tell you things you could have Googled
Accuracy: About as reliable as weather forecasts

The Golden Age (After Our Spy)

Research Time: 30 minutes while you grab coffee
Data Quality: Actually comprehensive and current (shocking, we know)
Cost: Whatever Runner H costs (probably less than your lunch budget)
Accuracy: Embarrassingly better than most humans

The Math That Matters

Time Saved: 95% (use the extra time for literally anything else)
Money Saved: 99% (your accountant will send thank-you cards)
Data Quality: 300% improvement (your competitors won't know what hit them)
Job Security: Significantly improved (you're welcome)

Real Talk: Who Actually Benefits

Small Businesses: Finally get intelligence without selling a kidney
Consultants: Deliver faster results and look like wizards
Product Managers: Stop guessing what competitors are doing
Anyone with Competitors: Which is... everyone with a business

This system basically democratizes competitive intelligence, making it accessible to anyone who can copy-paste a prompt. It's like giving everyone a Ferrari when they were used to walking.

Full Disclosure

The Truth Behind the Spy Operation:

Let's be honest here - this wasn't a "first try, perfect success" situation. Oh no, that would be far too easy and wouldn't involve nearly enough existential questioning.

The Real Development Story:

AI Assistant Used: Claude (Anthropic's AI) was my co-conspirator in this prompt engineering adventure
Iteration Count: 15+ versions of this prompt died horrible deaths before this one finally decided to cooperate
Debugging Sessions: Too many to count (Google Sheets refused to populate for the longest time, like a stubborn teenager)
Coffee Consumed: Classified information
Sanity Remaining: Questionable

What Went Wrong (A Lot):

Early versions created beautiful empty spreadsheets (very zen, completely useless)
Some versions generated HTML emails that made Gmail cry with formatting errors
Several attempts got stuck in "Phase 3" limbo like a bureaucratic nightmare
Multiple emoji-related casualties (turns out Runner H can be sensitive about special characters)
One version tried to set up continuous surveillance and probably would have achieved sentience

The Claude Collaboration:
This prompt is the result of an iterative development process where Claude helped refine the workflow, debug Runner H limitations, and optimize for the platform's specific syntax requirements. Think of it as AI helping AI help humans spy on other humans' businesses. Meta? Absolutely. Effective? You bet.

Lessons Learned:

Runner H doesn't like stopping mid-workflow (it gets commitment issues)
Google Sheets population requires very explicit instructions
Email formatting should be kept simpler than your dating profile
Sometimes the 16th time really is the charm

So yes, this "overnight success" took about 15 nights of prompt engineering. But hey, now YOU get to use the perfected version without going through the digital equivalent of software development hell.

Transparency: because pretending you're a prompt engineering genius on the first try is for amateurs.

Social Love

The Awkward Truth:

Listen, I built a digital spy that can infiltrate entire industries and generate comprehensive intelligence reports, but asking me to promote it on social media? That's where I draw the line.

I'm about as comfortable with self-promotion as a vampire is with garlic festivals. The irony isn't lost on me - I can automate corporate espionage but can't bring myself to post about it on LinkedIn without breaking into a cold sweat.

So, uh, if you found this useful and want to share it, that's cool I guess? I'll just be over here in my corner, proud but quietly mortified by any attention this might receive.