SafeDose AI - RegimenGuard

💡 Inspiration Managing multiple medications shouldn't require patients to become pharmacology experts. A single new medicine can introduce an interaction that's easy to miss. We built SafeDose AI — RegimenGuard to turn complex medication information into understandable safety insights — and with WebMCP, allow AI agents to work directly with structured medication context.

The agent investigates. SafeDose provides context. The human decides.

🛡️ What it does

SafeDose is an agent-native medication safety workspace that helps users:

💊 Manage their current medication regimen

⚠️ Detect potential drug interactions

📊 Understand an overall regimen safety score

🔴 Distinguish critical risks from cautions and safer combinations

🔎 Search structured medication information

👥 Support caregiver and medication-management workflows

🤖 Powered by WebMCP

SafeDose exposes three focused, read-only agent tools: search_medication 🔎 — Find medication information get_current_regimen 💊 — Understand the user's current regimen check_regimen_safety 🛡️ — Check a candidate medication against that regimen

Instead of repeatedly copying medication lists into an AI conversation, compatible agents can work with SafeDose's structured capabilities directly.

Agents investigate — they don't prescribe or modify medications.

🛠️ How we built it

SafeDose combines: React + TypeScript ⚛️ → application
WebMCP 🤖 → agent-native tools
Medication/FDA data 💊 → drug information
Google AI Studio ✨ → AI-assisted development
GitHub 🐙 → source control
Vercel ▲ → live deployment

We also built an Agent Activity interface to make WebMCP capabilities visible and understandable.

🧩 Challenges we ran into

Our biggest challenge was bridging two worlds: Human UI ↔ WebMCP ↔ AI Agent We worked through evolving WebMCP browser APIs, tool registration, deployment, agent discovery, and medication-data integration.

Healthcare also raised an important design question: How much authority should an AI agent have? Our answer was deliberate — read-only tools and human-controlled decisions.

🏆 Accomplishments that we're proud of

We're especially proud that SafeDose isn't simply “another healthcare chatbot.” We transformed an existing medication-safety experience into an agent-native web application, while preserving a polished interface for traditional users.

Most importantly, WebMCP addresses a genuinely contextual problem:

It's not only “What does this medication do?”
It's “What does this medication mean alongside everything I'm already taking?”

🧠 What we learned

WebMCP showed us that the future of AI on the web isn't necessarily putting a chatbot on every website.

It's about making websites themselves understandable and actionable to agents. And in healthcare, powerful AI doesn't always mean maximum autonomy. Good agents also know where their authority ends.

🚀 What's next for SafeDose AI — RegimenGuard

Next, we envision: 💊 Medication reconciliation
👨‍⚕️ Pharmacist-review workflows
👨‍👩‍👧 Caregiver collaboration
📅 Smarter medication planning
♿ Improved accessibility
🤖 Expanded WebMCP capabilities with human approval

Our goal is simple: