Her

• 

• 

• 

• 

• 

• 

• 

• 

• 

• 

• 

• 

• 

• 

• 

Inspiration

HER started with a simple observation: we overthink everything.

A text from someone can become a 30-minute investigation. A small decision can turn into ten different scenarios in our heads. Sometimes we don't need another search result—we need something that understands the context behind the situation.

I wanted to build something that felt less like a generic AI chatbot and more like a personal companion: something that could remember what matters, understand messy situations, and help us think through choices.

That became HER — “Let HER judge.”

The WebMCP Challenge gave me an opportunity to take that idea further by exploring what happens when a personal AI experience isn't limited to a human typing into a chat box, but can also interact with AI agents through structured web tools.

What it does

HER is a personal AI companion built for the things you overthink.

💌 KNOW ME

Users can share information about themselves, their preferences, experiences, and context. HER can use that context to make future interactions more personal.

⚖️ HER HONOR

Users bring HER a messy situation and let her judge.

Instead of simply generating a generic response, HER considers the context provided by the user and gives a clear, structured perspective on the situation.

🔮 WHAT IF

Users can explore different choices and ask, “What if I do this instead?”

HER helps compare possible paths and think through potential outcomes before making a decision.

🤝 Human + Agent Interaction

For the WebMCP Challenge, HER was extended with WebMCP capabilities so AI agents can interact with the application through structured tools rather than relying only on visual clicking or unstructured text.

This creates a different interaction model:

Human → HER → Agent → HER → Human

The human stays in control of the decision, while the agent can use HER's capabilities as part of a larger workflow.

How we built it

HER was built with a modern web stack:

• Next.js
• React
• TypeScript
• Tailwind CSS
• OpenAI
• WebMCP

The AI experience uses different model capabilities for different tasks, including lightweight context processing and more complex reasoning.

The most important addition for this challenge was WebMCP. Instead of treating HER as only a page that an agent can look at, the application exposes structured capabilities that an AI agent can discover and invoke.

The interface itself was designed around a soft, playful scrapbook/stationery aesthetic because I wanted the experience to feel personal and approachable rather than like another corporate AI dashboard.

Challenges we ran into

One of the biggest challenges was balancing personality with usefulness.

HER needed to feel like HER without becoming just a cute chatbot. Every interaction needed to have an actual purpose: remembering context, evaluating a situation, or helping someone explore a decision.

Another challenge was thinking about the application from both sides of the interaction. Designing for a human user is very different from designing capabilities that an AI agent can reliably discover and use.

WebMCP pushed me to think about functionality as structured capabilities rather than just UI elements.

I also had to make sure that the AI remained a supporting layer rather than taking control away from the user. HER can provide a perspective, but the final decision always belongs to the human.

Accomplishments that we're proud of

I'm proud that HER became more than a traditional AI chat interface.

The biggest accomplishment was connecting the original idea of a personal AI companion with the emerging idea of agent-accessible web applications.

I'm especially proud of:

• Creating a distinct identity and interaction model for HER.
• Building three complementary experiences: KNOW ME, HER HONOR, and WHAT IF.

Built With

• next
• react