PrepAgent

AI-powered interview research. Enter a company and interviewer — get a personalized briefing in 90 seconds.

What it does

PrepAgent runs a 4-agent research pipeline when you have an interview:

1. Company Intelligence Agent — finds recent news, funding, engineering blog posts, and job posting patterns
2. Interviewer Research Agent — builds a profile of who you're meeting and finds overlap with your background
3. RAG Retrieval Agent — queries your indexed resume and past briefings for relevant context
4. Synthesis Agent — generates a personalized briefing with talking points specific to your background

The output is a 4-section briefing: company intelligence, interviewer profile, your talking points, and questions to ask.

Tech Stack

Frontend: Next.js, TypeScript, Tailwind CSS, Supabase Auth

Backend: FastAPI, Python 3.12, LangGraph, LangChain, Claude Sonnet

Database: Supabase PostgreSQL

Vector Store: Pinecone

Reranking: Cohere Rerank

Observability: LangSmith

Evals: RAGAS

Integrations: Gmail MCP, Google Calendar MCP, Google Drive MCP

Infrastructure: Docker, AWS EC2, GitHub Actions CI/CD

Architecture

Next.js Frontend (Vercel)
    ↓
FastAPI Backend (AWS EC2)
    ↓
LangGraph Supervisor Graph
    ├── Company Intelligence Agent (Tavily)
    ├── Interviewer Research Agent (Tavily)
    ├── RAG Retrieval Agent (Pinecone + Cohere Rerank)
    └── Synthesis Agent (Claude Sonnet)
    ↓
Gmail MCP + Google Calendar MCP + Google Drive MCP
Supabase PostgreSQL
LangSmith Tracing

Getting Started

Prerequisites

• Python 3.12+
• Node.js 18+
• Accounts: Anthropic, OpenAI, Cohere, Pinecone, Supabase, LangSmith, Tavily

Backend Setup

cd backend
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
# fill in your API keys
uvicorn main:app --reload

Frontend Setup

cd frontend
npm install
cp .env.local.example .env.local
# fill in your Supabase keys
npm run dev

Environment Variables

backend/.env

SUPABASE_URL=
SUPABASE_SERVICE_KEY=
ANTHROPIC_API_KEY=
OPENAI_API_KEY=
COHERE_API_KEY=
PINECONE_API_KEY=
PINECONE_INDEX=prepagent
LANGSMITH_API_KEY=
LANGSMITH_PROJECT=prepagent
TAVILY_API_KEY=
GOOGLE_CLIENT_ID=
GOOGLE_CLIENT_SECRET=
FRONTEND_URL=http://localhost:3000

frontend/.env.local

NEXT_PUBLIC_SUPABASE_URL=
NEXT_PUBLIC_SUPABASE_ANON_KEY=
NEXT_PUBLIC_API_URL=http://localhost:8000

Agent Architecture

PrepAgent uses a LangGraph supervisor graph with stateful routing across 4 specialized agents. The supervisor decides when each agent has gathered sufficient context before handing off to synthesis.

Each agent step is traced in LangSmith. Every tool call is logged with token usage and latency.

Evaluation

PrepAgent uses RAGAS to evaluate briefing quality across 3 metrics:

• Answer relevance — does the briefing address this specific role and company
• Faithfulness — are talking points grounded in what the research actually found
• Context precision — is the retrieved context from Pinecone useful for this briefing

Baseline scores are documented and tracked across prompt iterations.

License

MIT