CampusPilot — Agentic Campus OS

A WebMCP-powered campus agent that lets students and AI agents discover spaces, check availability, plan tasks, and execute campus actions together—with humans always in control.

Comment

Inspiration

Campus life involves many small tasks that become surprisingly difficult when information is scattered across different systems. Students may need to find a quiet study space, check availability, plan their day, or perform campus-related actions, but they often have to manually search through multiple pages and interfaces.

We wanted to explore a different approach: what if a campus website could be used directly by an AI agent as a set of structured tools, while still keeping the student in control?

That idea led to CampusPilot — Agentic Campus OS, a WebMCP-powered campus assistant designed to connect students and AI agents with useful campus actions.

What it does

CampusPilot provides an agent-friendly interface for common campus workflows such as:

  • 🔎 Discovering available campus spaces
  • 📚 Finding suitable study locations
  • 🗓️ Planning campus-related tasks
  • ⚡ Executing actions through structured WebMCP tools
  • 🤖 Showing an agent's reasoning and planned actions
  • 👤 Keeping the human involved through approval before important actions

Instead of treating the website as something an agent can only read, CampusPilot exposes functionality through WebMCP tools that an AI agent can understand and invoke.

How we built it

CampusPilot was built as a modern React web application with a premium, responsive interface.

The core architecture is organized around a WebMCP tool layer. We created structured tools with defined schemas and handlers so that an agent can interact with campus functionality in a predictable way.

The application includes:

  • React + TypeScript for the application
  • Tailwind CSS for the interface
  • Lucide icons for UI elements
  • WebMCP imperative APIs for agent-tool interaction
  • A deterministic demo agent for demonstrating the complete workflow
  • An Agent Reasoning timeline to make agent actions understandable
  • A Plan Card that allows the human to review and approve planned actions

The most important part of the implementation is that the demo agent and the WebMCP interface use the same underlying tool logic. This helped us demonstrate how an AI agent could move from understanding a user's request to selecting tools and executing a campus workflow.

What we learned

The biggest lesson was that building an agentic application is not only about adding an AI model.

The interface between the agent and the application matters just as much.

WebMCP provides a structured way to expose website capabilities to agents. Designing clear tool names, schemas, inputs, outputs, and handlers made us think about the application from an agent's perspective rather than only from a human user's perspective.

We also learned that human-in-the-loop interaction is important. An agent should be able to plan and recommend actions quickly, while the user should still understand what is going to happen before approving important operations.

Challenges we faced

One of the main challenges was designing WebMCP tools that were both useful for humans and understandable to an AI agent.

We had to carefully structure tool inputs and outputs so that each action had a clear purpose and predictable behavior.

Another challenge was creating a convincing agent experience without hiding what the agent was doing. Instead of simply showing a final result, CampusPilot exposes an Agent Reasoning timeline and planned actions so users can follow the workflow.

We also had to ensure that the application remained reliable during the demo. We tested the application build and the complete tool interaction flow to make sure the WebMCP functionality worked consistently.

Why CampusPilot?

CampusPilot is a small step toward a future where websites are not just pages that humans navigate, but interactive environments that AI agents can safely operate through structured capabilities.

Our goal is to extend this concept beyond a demo into a real campus operating layer where students can ask an agent to accomplish tasks across campus systems while remaining in control of the final decisions.

CampusPilot turns a campus website into an agent-ready environment — connecting human intent, WebMCP tools, and AI agents in one workflow.

Built With

Share this project:

Updates

Submission history