This repository contains code examples for building agentic AI workflows using OpenRouter API, which provides access to multiple AI models (OpenAI, Anthropic, Google, etc.) through a single, consistent interface.

• Access models from multiple providers through one unified API
• Compare responses from different AI models
• Build multi-step reasoning agents that break down complex tasks
• Example workflow for research and recommendation tasks

• Python 3.8 or higher
• OpenRouter account (sign up at openrouter.ai)

1. Clone this repository:

git clone https://github.com/your-username/agentic-ai-workflow
cd agentic-ai-workflow

2. Create a virtual environment:

python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Create a .env file with your API keys:

OPENROUTER_API_KEY=your_openrouter_api_key_here
YOUR_SITE_URL=https://yourdomain.com
YOUR_SITE_NAME=Your App Name

Test your OpenRouter API connection:

python basic_setup.py

Compare responses from different AI models:

python model_comparison.py

Execute a multi-step reasoning workflow:

python agent_example.py

• basic_setup.py - Basic OpenRouter API client setup
• model_comparison.py - Compare responses from different models
• agent_example.py - Complete agentic workflow implementation
• requirements.txt - Project dependencies

• Add vector database for long-term memory
• Enable external tools (web search, calculators, etc.)
• Implement model routing for cost optimization
• Add retry logic and better error handling
• Implement user feedback loops

For a detailed tutorial on building agentic AI workflows with OpenRouter API, check out the article:

• Building Your First Agentic AI Workflow with OpenRouter API by Allan Niñal

MIT