Clara

AI Calendar Agent — Google Calendar Management with Orgo + OpenAI + Anthropic

Comment

🧠 AI Calendar Agent — Google Calendar Management with Orgo + OpenAI + Anthropic

This project integrates Orgo, OpenAI, and Anthropic to create an autonomous agent that can manage your Google Calendar through natural language prompts. It uses Orgo’s virtual desktop infrastructure, Agent S2 for GUI automation, and large language models for reasoning and control.

🚀 Features

  • ✅ Create, edit, or delete calendar events using natural language
  • 🗓️ View and search for existing meetings or appointments
  • 🧭 Navigate Google Calendar in a Chrome browser GUI
  • 🔄 Switch between OpenAI (GPT-4o) or Anthropic (Claude) as the control model
  • 🖥️ Works in Orgo cloud VMs or your local machine

📦 Tech Stack

Component Description
Orgo GUI environment and action execution
Agent S2 Looping agent framework (perception → LLM → act)
OpenAI API GPT-4o reasoning and planning
Anthropic API Claude as an alternative controller model
Google Calendar (web) Controlled via Chrome in a GUI environment

🔧 Setup

1. Clone the Repo

git clone https://github.com/your-org/ai-calendar-agent.git
cd ai-calendar-agent

2. Set Up Environment

install vscode extension to use ipynb Open anaconda prompt and make your current directy that of this repository

conda env create -f orgo-env.yml
conda activate orgo-env

Select python kernel of the ipynb to be orgo-env

Alternatively, if you prefer using .py files, run the python file through anaconda prompt, like a normal python file

3. Configure API Keys

Edit .env with your credentials:

ORGO_API_KEY=your_orgo_key
OPENAI_API_KEY=your_openai_key
ANTHROPIC_API_KEY=your_anthropic_key
BROWSER_EXECUTABLE_PATH=/path/to/chrome

4. Launch Agent

python run_agent.py

The agent will:

  • Launch an Orgo virtual desktop (or local desktop)
  • Open Chrome and navigate to Google Calendar
  • Accept your natural language prompt
  • Perform calendar tasks accordingly

🗣️ Example Prompts

"Create a meeting tomorrow at 3pm with Sarah about Q3 planning."
"Find my next event with John."
"Reschedule my 10am call to Friday."
"Cancel the 'Yoga' event on Saturday."

🛠️ Agent Behavior

Agent S2 follows a loop:

  1. Takes a screenshot of the screen
  2. Sends it to the selected LLM (OpenAI or Anthropic)
  3. LLM determines next GUI action
  4. Executes mouse/keyboard commands via Orgo API
  5. Repeats until task is completed

⚙️ Configuration

You can tweak behavior in config.yaml:

model: openai  # or 'anthropic'
max_steps: 30
temperature: 0.7
browser: chrome
calendar_url: "https://calendar.google.com"

📁 File Structure

.
├── run_agent.py         # Main loop for launching and controlling the agent
├── actions/             # Mouse/keyboard control logic
├── prompts/             # Prompt engineering templates
├── config.yaml          # Configuration file
├── .env                 # API keys and local paths
├── requirements.txt     # Python dependencies
└── README.md            # This file

📌 Notes

  • You must be signed into Google Calendar in the browser for automation to work.
  • Security: Keep your API keys safe. Do not hardcode them.
  • You can switch between OpenAI and Anthropic via config or .env.

📎 References

🤝 Contributions

Pull requests are welcome! For major changes, open an issue first to discuss your idea.

👥 Team

Built by: @nibess1 @weisintai @darriusnjh @SaaiAravindhRaja

Built With

Share this project:

Updates