optix-ai

The open-source FinOps agent that makes your cloud spend smarter.

optix-ai connects to AWS Cost Explorer and uses an LLM (Claude or Gemini) to generate prioritized, actionable cost optimization reports — covering service breakdowns, spending trends, EC2 rightsizing, Savings Plans, and anomaly detection.

Features

• Cost breakdown — top services by spend over the last 30 days
• Trend analysis — month-over-month changes for the last 3 months
• EC2 rightsizing — identify oversized instances and estimated savings
• Savings Plans — committed-use discount recommendations
• Anomaly detection — unexpected cost spikes above a configurable threshold
• Multi-provider — supports Anthropic Claude (default) and Google Gemini
• Report export — save the full report as a Markdown file

Requirements

• Python 3.11+
• AWS credentials with Cost Explorer access
• Anthropic API key (or Google API key for Gemini)

Installation

git clone https://github.com/techtious/optix-ai.git
cd optix-ai
pip install -e .

For Gemini support:

pip install -e ".[gemini]"

Configuration

Copy .env.example to .env and fill in your credentials:

cp .env.example .env

# Required
ANTHROPIC_API_KEY=sk-ant-...

# AWS credentials (or use AWS_PROFILE / instance role)
AWS_ACCESS_KEY_ID=
AWS_SECRET_ACCESS_KEY=
AWS_DEFAULT_REGION=us-east-1

# Required only for --provider gemini
GOOGLE_API_KEY=

AWS credentials can also be provided via AWS_PROFILE or an instance role — any method supported by the AWS SDK works.

Usage

# Generate and display a report (uses Claude by default)
optix

# Use Gemini instead
optix --provider gemini

# Save report to a Markdown file
optix --save

# Save to a specific directory without printing to terminal
optix --save ./reports --no-display

Report structure

Each generated report includes:

1. Executive Summary — total spend, total identified savings, top 3 actions
2. Current Cost Landscape — service breakdown table with month-over-month trends
3. Optimization Recommendations — ranked by estimated annual savings
4. Detected Anomalies — unexpected cost spikes with impact amounts
5. Next Steps — numbered action plan

License

MIT