What's next for Syndassist

Syndassist: AI-Powered Networking That Grows With You

🌱 The Inspiration

Networking is exhausting. I spent hours researching people on LinkedIn, crafting personalized messages, and then—worst of all—forgetting to follow up. Connections would go cold, opportunities would slip away.

I wanted networking to feel natural, rewarding, and even fun. What if your network was a garden? What if AI handled the research while you focused on genuine relationships?

That question became Syndassist.

🎯 What It Does

Syndassist automates networking while keeping it personal:

Discovery Agent: Takes custom queries like "Find biotech VCs in SF" and discovers relevant professionals with relevance scores
Outreach Agent: Generates personalized messages using web research + your past interaction notes
Journal System: AI summarizes your coffee chats, then references those details in future outreach
Event Discovery: Finds networking events and suggests who to contact beforehand
Garden Gamification: Your network grows like a plant—the more you nurture connections, the healthier your garden

🛠️ How I Built It

Stack: React + TypeScript, FastAPI, Dedalus Labs, Supabase

Architecture:

React Frontend → FastAPI → Dedalus Runner → Claude AI → Supabase

Key Features:

Natural language discovery ("Find AI researchers at Stanford")
Real-time streaming results
Journal-powered message personalization
4 specialized AI agents working together
Nature-themed UI with earth tones and plant animations

🚧 Biggest Challenges

**1.Integrating multiple external systems – Connecting Supabase Auth, Dedalus agents, Brave Search MCP, and your own backend without leaks, auth issues, or inconsistent state. Each has its own errors, rate limits, and behaviors.

**2. Keeping AI responses reliable – Keeping the discovery/event agents working when Brave Search is rate limited, and handling cases where the model returns non-JSON or error text instead of structured data.

**3. Production readiness – Managing env vars, CORS, ports, and deployment across frontend/backend (e.g. Vercel + Railway) so the app works end-to-end outside local development.

—

📚 What I Learned

**Graceful degradation – When external APIs fail (e.g. Brave Search rate limits), fall back to the model’s own knowledge or cached-style data instead of failing completely.
**Credentials ≠ ownership – Passing your Brave API key to a hosted MCP doesn’t guarantee your account sees usage; hosted services often use their own keys behind the scenes.
**Clear error handling end-to-end – Surface backend errors (e.g. rate limits, onboarding failures) in the UI with actions (e.g. “Try again in a few minutes”) instead of generic “Something went wrong.”

💡 What I'm Proud Of

The UX feels smooth and intuitive, with contacts appearing in real time and the garden growing as your network does. The full stack actually works end to end (React → FastAPI → Dedalus → Claude → Supabase) I’m using it for my own networking now, not just as a demo. Built in one intense day of coding, debugging, and many “why isn’t this working” moments.