First Founders

Majority of first-time founders fail because they didn't have anyone to show them the path. First Founders is an AI mentor that guides you from idea to investor-ready- step by step.

Comment

Inspiration

We both felt the pain of hitting a wall after coming up with an amazing business idea. Bobbie started a business, had her idea taken, and had no idea how to find investors or validate her market. Jaliyah faced the same overwhelming silence most first-time founders encounter: no roadmap, no mentor, no clear next step. We built First Founders because it's the tool we both needed and didn't have.

What it does

First Founders is an AI-powered mentor that guides first-time founders from idea to investor-ready- step by step. You describe your idea. The agent validates your market with real data, protects your concept before you pitch it to anyone, generates your customer discovery questions, walks you through legal setup by country, and scores your investor readiness across five critical areas. The agent thinks across three modes. Innovative (obsessed with simplicity and what makes something great), Strategic (first principles, challenge every assumption), and Intuitive (market reality, data over opinion) blending into one coherent mentor voice guided by the YC framework.

How we built it

Agent architecture:

  • Google ADK — we built a sub-agent architecture because ADK doesn't allow google_search and custom tools in the same agent. A dedicated search_agent handles all market research via Google Search, wrapped as an AgentTool on the main root_agent.
  • Gemini 2.5 Flash via Vertex AI — provides reasoning, feedback, and content generation.
  • MongoDB MCP Server — integrated MCPToolset directly into the ADK agent, it has raw database access via find, aggregate, and listCollections beyond our pre-written Python tools. Required installing Node 22 and mongodb-mcp-server globally in the Docker image to work in production. Database:
  • MongoDB Atlas — three collections: founders , notes, resources
  • tools.py — six custom Python tools: save_founder, get_founder, update_milestone, save_note, save_resource, get_notes_and_resources. Each is a typed function the ADK agent calls directly.
  • Investor score algorithm — 0–100 score across five areas (problem clarity, market size, differentiation, traction signals, founder-market fit). Auto-recalculates every time a milestone completes..
  • PIN security — SHA-256 hashed PINs stored in MongoDB. New users set a PIN at the end of onboarding. Returning users get a PIN keypad modal in the sidebar. Backend:

Frontend:

  • Vanilla JS + CSS

Deployment:

  • Google Cloud Run — containerized with a custom Dockerfile.

Challenges we ran into

  • Vertex AI authentication — gcloud credential corruption on Windows mid-build required a full reinstall of Google Cloud SDK. Then migrated from AI Studio keys to Vertex AI with proper ADC and quota project setup.
  • MCP + ADK tool conflict — ADK raises an error if google_search and custom Python tools are in the same agent. Solved by wrapping search in a dedicated sub-agent using AgentTool.

Accomplishments that we're proud of

  • MongoDB MCP Server running in production inside a Cloud Run container.
  • Prompt engineering a mentor personality that feel experienced, direct, and warm.
  • Genuinely working hard and having together.

What we learned

  • Prompt engineering is product engineering
  • MCP Servers are a great asset
  • Vertex AI credentials require more setup than AI Studio keys

What's next for First Founders

  • Document generation — export pitch decks, business summaries, and investor one-pagers directly from the conversation
  • Founder matching — connect founders working on similar problems or complementary ideas based on their profiles and milestones
Share this project:

Updates