This project creates an AI agent from scratch using Node.js and Google's Gemini API, based on the open-source tutorial from Hendrixer/agent-from-scratch. It replaces OpenAI with Google Gemini, making it completely free to experiment with.

✅ Perfect for developers looking to explore AI agents without incurring any costs.

• Uses Google Gemini API (@google/genai)
• No OpenAI key needed – works entirely on Gemini’s free tier
• Built in TypeScript for type safety and modular design
• Based on a proven educational template
• Fully extensible: add tools, memory, workflows, and more

• Node.js v18+
• A free Google Gemini API key from AI Studio

👉 Get your Gemini API key here: Google AI Studio API Key

1. Clone the repository

git clone https://github.com/gsk-007/ai-agent-gemini.git
cd ai-agent-gemini

2. Install dependencies

npm install
# or
bun install

3. Add your environment variables

Create a .env file in the root directory with the following content:

GEMINI_API_KEY=your_google_api_key_here

🔐 Keep your API key secret! Never commit it to version control.

4. Chat With Model

npm run dev "<your message>"
# or
bun run index.ts "<your message>"

├── src/
│   ├── agent.ts          # Core agent logic using Gemini
│   ├── tools.ts          # Extendable tool interface
│   ├── memory.ts         # Optional memory implementation
│   └── types.ts          # Shared types and interfaces
├── index.ts              # Entry point
├── .env                  # Environment variables (not committed)
├── package.json
└── README.md

Here’s what you get for free:

• 15 requests/minute
• 1,500 requests/day
• 1 million tokens/minute
• Cost: Free

• 2 requests/minute
• 50 requests/day
• 32k tokens/minute
• Cost: Free

⚠️ Limits are generous for prototyping but can be raised with billing enabled.

More info: https://ai.google.dev/pricing

• Original Agent Project (Hendrixer)
• Gemini SDK (docs)
• Google AI Studio

MIT – Use it, modify it, build your own agent!

PRs welcome! Feel free to fork the project, add your own tools, or extend the agent's reasoning abilities.