DueDeck

Inspiration

We've watched countless founders spend weeks perfecting pitch decks while VCs waste 10-20 hours per deal on technical due diligence. Meanwhile, 65% of deals die because investors can't assess technical risk fast enough. We asked ourselves: What if AI could do the heavy lifting for both sides? During YC office hours, a founder told us they spent 40 hours creating materials for a single investor meeting, only to be rejected in 15 minutes because the VC "didn't understand the tech stack." That conversation sparked DueDeck. We realized GitHub repos contain everything investors need, but nobody has time to analyze them properly.

What it does

DueDeck transforms any GitHub repository into a complete investor package in under 5 minutes:

Technical Due Diligence Report (8-12 pages) - Enterprise-grade analysis covering architecture quality, security vulnerabilities, scalability assessment, technical debt quantification, and team code patterns [add another feature here]

Target users: VCs conducting Series A-B diligence, founders preparing for fundraising, and accelerators evaluating applications.

How we built it

Frontend & User Experience

Next.js 14 with App Router for server-side rendering and optimal performance
Figma Design and Figma Make for UI/UX Design
Tailwind CSS + shadcn/ui for a polished, investor-grade interface
React Query for smart data fetching and caching -Coderabbit for GitHub management
MCP Automation as the central nervous system coordinating all specialized agents
Letta agents with memory architecture that maintain context across 1000+ file repositories
CodeRabbit for deep code review, security scanning, and technical debt assessment
Bright Data web scraping infrastructure: API & Integration Hub
PostgreSQL + Prisma ORM for relational data
Redis for job queuing and caching
NextAuth.js with GitHub OAuth

Challenges we ran into

There was a limit the number of accounts we could scrape for data. Connecting the MCP servers. Battery life of laptops and wifi.

Accomplishments that we're proud of

Connecting the MCP servers, able to find a workaround the web scraping, having a finishing product that can benefit people in the builder space.

What we learned

MCP is incredibly powerful for orchestrating multiple AI agents. It's like having a conductor for an AI orchestra Elastic's semantic search + Letta's memory = game-changer for understanding massive codebases CodeRabbit's API provides more nuanced insights than we expected (detected anti-patterns we didn't even know to look for) Bright Data's proxy network is essential for gathering competitive intelligence at scale without getting blocked

What's next for DueDeck

We want to connect remaining mock integrations to live APIs and conduct user testing with 5-10 founders Implement report customization (let users hide/emphasize specific metrics) Build comparison mode to analyze your repo against competitor repositories

If we continue building:

Custom report templates for different investor types (angels vs. VCs vs. corporate acquirers)
Team collaboration features (comments, shared workspaces)
Historical tracking to show code quality evolution over time
Public API for integration with existing VC deal flow tools