Autonomous FinOps Agent: From Cloud Analysis to Cost Actions

An AI-powered FinOps agent that detects cloud cost anomalies, recommends optimizations, and automatically creates GitLab issues for engineering follow-up.

What it does

• Reads mock GCP billing data at SKU/project/day level
• Detects unusual cloud cost spikes
• Suggests optimization actions
• Creates GitLab issues automatically
• Provides a deployed API that Gemini can call as an agent tool

Architecture

Gemini Agent
   ↓
Cloud Run FastAPI backend
   ↓
Mock GCP billing CSV
   ↓
Anomaly detection + recommendations
   ↓
GitLab issue creation

Tech Stack

• Python
• FastAPI
• Google Cloud Run
• Google Secret Manager
• GitLab API
• Gemini API
• Pandas

Project Structure

.
├── main.py
├── raw_gitlab.py
├── agent.py
├── gcp_mock_billing_data.csv
├── requirements.txt
├── Dockerfile
├── .gitignore
└── README.md

API Endpoints

Health check

GET /

Full FinOps report

GET /full_finops_report

Create GitLab ticket

POST /create_ticket

Example body:

{
  "title": "FinOps Test Ticket",
  "description": "Created from Cloud Run API",
  "priority": "high"
}

Run autonomous flow

GET /run_finops_agent

This runs:

analyze costs → detect anomalies → create GitLab issues

Local Setup

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --reload --port 8080

Open:

http://localhost:8080/docs

Google Cloud Setup

Enable:

• Cloud Run Admin API
• Cloud Build API
• Artifact Registry API
• Secret Manager API

Create a GCP secret named:

gitlab-token

The Cloud Run service account needs:

Secret Manager Secret Accessor

Deploy:

gcloud run deploy finops-agent \
  --source . \
  --region europe-west1 \
  --allow-unauthenticated

Google Cloud Configuration

Before running the application, update the GCP project ID in raw_gitlab.py:

project_id = "YOUR_GCP_PROJECT_ID"

Replace this value with your Google Cloud project ID.

The application uses this project to retrieve the GitLab token from Google Secret Manager.

GitLab Setup

Create a GitLab personal access token with api scope.

The app currently uses this project ID:

GITLAB_PROJECT_ID = "YOUR_GITLAB_PROJECT_ID"

Update it in main.py if you use another repo.

Agent Configuration

The Gemini agent is implemented in agent.py.

Before running the agent, update the following configuration values:

# Gemini API key from Google AI Studio
GEMINI_API_KEY = "YOUR_GEMINI_API_KEY"

# Deployed Cloud Run API URL
FINOPS_API = "https://YOUR_CLOUD_RUN_URL"

Step 1: Create a Gemini API Key

1. Open Google AI Studio.
2. Generate a Gemini API key.
3. Replace:

GEMINI_API_KEY = "YOUR_GEMINI_API_KEY"

with your own API key.

Step 2: Deploy the FinOps API

Deploy the FastAPI backend to Google Cloud Run.

After deployment, Cloud Run will provide a service URL similar to:

https://finops-agent-xxxxxxxxxx-ew.a.run.app

Replace:

FINOPS_API = "https://YOUR_CLOUD_RUN_URL"

with your deployed endpoint.

Step 3: Run the Agent

python agent.py

The agent will:

1. Retrieve cloud cost analysis from the FinOps API.
2. Analyze anomalies using Gemini.
3. Generate recommendations.
4. Create actionable GitLab remediation tickets through the deployed API.

Example Workflow

Gemini Agent
      ↓
Cloud Run FinOps API
      ↓
Analyze Billing Data
      ↓
Detect Anomalies
      ↓
Generate Recommendations
      ↓
Create GitLab Issue

Summary

Autonomous FinOps Agent closes the loop between cloud cost visibility and engineering action. Instead of only reporting spend anomalies, it detects issues, explains business impact, and creates actionable GitLab tickets for teams to fix them.

(Configurations to replace with your parameters: YOUR_GCP_PROJECT_ID, YOUR_GITLAB_PROJECT_ID, YOUR_GEMINI_API_KEY, YOUR_CLOUD_RUN_URL)

Disclaimer

This project uses mock GCP billing data for hackathon demonstration purposes.