Skip to content

SaharPak/founder-school-agent

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Founder School Agent

An AI agent that helps founders move from idea to action to event.

Built by Sahar Pakseresht as the application project for the Ecosystem & Events Coordinator role at Aalto University Founder School.

Made with Node.js Express Runs offline License: MIT

Free to try, no login required. Hosted on Render's free tier, so the first load can take up to ~50 seconds to wake if it has been idle.

Overview · Features · Architecture · Run locally · Screenshots · Deploy


"If you are interested in this role, make it yours! Make use of your agency, ambition and energy to proactively produce something… Aalto Founder School is for people of action." — from the job posting

So instead of writing about what I would do, I shipped working software that does it.


🚀 Overview

Founder School Agent is a runnable AI web app that acts as a co-pilot for early-stage founders and the people who support them. It bundles three tools into one agent:

  1. A Founder Coach that gives bold, action-first advice.
  2. An Idea Validator that turns a raw idea into structured, honest feedback.
  3. An Event Architect that produces a complete, ready-to-ship plan for a Founder Talk or workshop in seconds.

It runs fully offline with a single command (no API key required) and seamlessly upgrades to live LLM answers the moment you add a key. The entire project is intentionally lightweight so a reviewer can clone it, run it, and understand it in under five minutes.

🎯 Problem Statement

Startup ecosystems run on two scarce resources: founder momentum and operator time.

  • Founders stall between having an idea and taking the next concrete step. They get generic advice, over-analyze, and lose weeks before talking to a single real user.
  • Ecosystem and events teams are small and fast-moving. Producing a high-quality event — topic, speaker brief, run-of-show, promo copy, success metrics — is repetitive, time-consuming work that eats into the time that should go toward people and relationships.

Both problems share a root cause: too much friction between intent and action.

💡 Solution

Founder School Agent removes that friction with a single AI agent that always pushes toward a concrete next step:

  • It coaches founders to act in the next 48 hours, not to ruminate.
  • It validates ideas with cheap experiments you can run this week, not vague encouragement.
  • It compresses hours of event production into seconds, giving a small team real leverage.

The design philosophy is idea → working software: ship something real, then let reality tell you what to build next.

✨ Features

Tool What it does Why it matters
💬 Founder Coach A conversational mentor in Founder School's bold, action-first voice. Streams its reply token by token and always ends with one concrete next step. Empowering future founders is the heart of the role.
🔍 Idea Validator Turns a raw idea into structured JSON: problem, audience, value proposition, top risks, a fit score, and cheap experiments to run this week. Helps the community move faster and fail cheaper.
🎤 Event Architect Generates a full event plan: title, tagline, speaker brief, run-of-show, promo kit (Instagram + email + poster), success metrics, and a production checklist. Directly mirrors the day-to-day of the events role.

Under the hood:

  • Zero-setup demo mode — a deterministic, on-brand engine means every feature works without any API key, account, or billing.
  • Provider-agnostic LLM layer — works with OpenAI, Anthropic (Claude), OpenRouter, Groq, Azure, or a local Ollama / LM Studio server. Same code, swap the env vars.
  • Production-shaped — API keys stay server-side, requests time out, and structured tools return validated JSON with a defensive fallback.
  • No build step — vanilla JS frontend styled with Aalto's real design tokens. Fast to read, run, and trust.

🧭 Example Use Cases

  • A first-time founder asks the Coach "I'm scared to launch, what do I do?" and gets a calm, motivating answer plus a single 48-hour action.
  • A student with a rough idea pastes it into the Validator and instantly sees the real problem, the right first audience, the biggest risks, and three experiments to run before Friday.
  • An events coordinator types "fundraising your first €100k" into the Event Architect and walks away with a complete Founder Talk plan — speaker brief, run-of-show, promo copy, and metrics — ready to publish.

More worked examples with real inputs and outputs live in docs/EXAMPLES.md.

🏗️ Architecture Overview

The agent follows a simple, robust pattern: every request tries the live LLM first and gracefully falls back to a deterministic on-brand engine, so the product is never broken even with no key.

flowchart TD
    U["👤 User<br/>(founder / events coordinator)"] --> APP

    subgraph APP["🧭 Founder School Agent (web app)"]
        UI["Single-page UI<br/>Coach · Validator · Event Architect"]
        API["Express API<br/>/api/chat · /api/validate · /api/event"]
        UI --> API
    end

    API --> ROUTER{"API key set?"}
    ROUTER -- "Yes" --> LLM["🤖 LLM / AI Agent<br/>OpenAI · Anthropic · Groq · local"]
    ROUTER -- "No" --> DEMO["⚙️ Deterministic demo engine<br/>(on-brand, offline)"]

    LLM --> REASON["🧠 Reasoning<br/>role-specific system prompts<br/>+ structured JSON tools"]
    DEMO --> REASON

    REASON --> OUT["✅ Recommendations & next actions<br/>coaching reply · validation report · event plan"]
    OUT --> U
Loading

For a deeper walkthrough of the request lifecycle, design decisions, and trade-offs, see docs/ARCHITECTURE.md.

🛠️ Tech Stack

Layer Choice Why
Runtime Node.js 20+ Native fetch, top-level await, built-in .env loading.
Server Express Minimal, well-understood, easy to read.
Frontend Vanilla JS + HTML + CSS No build step, no framework lock-in, instant to run.
AI Any OpenAI-compatible API Provider-agnostic; demo engine when no key is present.
Deploy Render / Hugging Face Spaces One-click blueprint included. See DEPLOYMENT.md.

📦 Installation

Requirements: Node.js 20+.

git clone https://github.com/SaharPak/founder-school-agent.git
cd founder-school-agent/agent
npm install

▶️ Running Locally

npm start
# → open http://localhost:3000

That's it. No API key required — the app runs in demo mode and every feature works.

To enable live LLM answers, add a key (any OpenAI-compatible provider):

# OpenAI
OPENAI_API_KEY=sk-... npm start

# or Anthropic (Claude) via its OpenAI-compatible endpoint
OPENAI_API_KEY=sk-ant-... \
OPENAI_BASE_URL=https://api.anthropic.com/v1 \
OPENAI_MODEL=claude-haiku-4-5-20251001 \
npm start

The header pill shows whether you're in Demo or Live mode. See agent/.env.example for all options.

📸 Screenshots

Home screen Founder Coach (live conversation)
Home screen AI response
Idea Validator Event Architect
Idea Validator Event Architect

Terminal — runs with one command, zero config:

Terminal execution

🔭 Future Improvements

  • Persist past coaching sessions and idea reports per user.
  • Export event plans to PDF / Notion / Google Calendar in one click.
  • A lightweight RAG layer grounded in Founder School's own resources and past events.
  • Analytics dashboard for the events team (attendance, NPS, ecosystem connections).

The full roadmap is in docs/FUTURE.md.

🗂️ Repository Structure

founder-school-agent/
├── README.md                ← you are here
├── DEMO.md                  ← 90-second demo video script
├── DEPLOYMENT.md            ← deploy to Render / Hugging Face Spaces
├── ELEVATOR_PITCH.md        ← 30-second spoken pitch
├── LINKEDIN_POST.md         ← ready-to-post project announcement
├── SHORT_DESCRIPTION.md     ← resume / portfolio blurb
├── LICENSE                  ← MIT
├── render.yaml              ← one-click Render deploy blueprint
├── docs/
│   ├── ARCHITECTURE.md      ← how the system works
│   ├── EXAMPLES.md          ← realistic prompts and outputs
│   ├── FUTURE.md            ← roadmap and future improvements
│   ├── demo/                ← place your recorded demo video here
│   └── screenshots/         ← app screenshots used in this README
└── agent/                   ← the AI agent (runnable web app)
    ├── server.js            ← Express app + 3 API routes (LLM → demo fallback)
    ├── lib/
    │   ├── llm.js           ← provider-agnostic OpenAI-compatible client
    │   └── demoEngine.js    ← deterministic on-brand responder (offline mode)
    ├── public/              ← single-page UI (HTML, CSS, JS, assets)
    ├── Dockerfile           ← container image (for Hugging Face Spaces)
    ├── .env.example         ← configuration template
    └── README.md            ← agent-level technical notes

📄 License

Released under the MIT License.

This is an independent application project. "Aalto", the Aalto logo, and "Founder School" are trademarks of Aalto University and are used here for context only.


Idea → working software. That's the loop I'd bring to the Founder School team.

Made with intent by Sahar Pakseresht

About

Founder School Agent - an AI agent that helps founders take action (Aalto Founder School application)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors