Autonomous FinOps Agent: From Cloud analysis to Cost Actions

An AI agent that doesn’t just analyze cloud costs — it detects issues, explains them, and takes action (tickets, alerts, optimizations) autonomously.

Comment

Inspiration

Cloud has transformed IT spending from predictable CAPEX to dynamic, usage-based OPEX — creating a new challenge: uncontrolled and complex cloud costs.

Finance teams are expected to act as strategic partners, yet they often rely on manual analysis and delayed reporting. Meanwhile, engineering teams lack real-time visibility into the financial impact of their decisions.

We asked: What if a FinOps analyst didn’t just analyze cloud costs — but actively fixed them?

What it does

Autonomous FinOps Agent: From Cloud Analysis to Cost Actions is an AI agent that detects cloud cost issues and takes action automatically.

The agent:

  • Monitors cloud cost data
  • Detects anomalies (e.g., sudden cost spikes)
  • Identifies root causes
  • Translates findings into business impact (€, %, unit cost)
  • Recommends optimizations
  • Automatically creates actionable tickets in GitLab

It also learns over time by storing past incidents and actions in MongoDB.

👉 Not just insights — execution.

How we built it

We built a tool-augmented autonomous agent powered by Gemini.

Core components:

  • Gemini → reasoning, planning, and decision-making
  • MCP server → structured tool interface
  • MongoDB → memory (incidents, actions, outcomes)
  • GitLab → execution (automated ticket creation)

Agent workflow:

Plan → Analyze → Detect → Decide → Act → Learn

Custom tools:

  • get_cloud_costs
  • detect_anomalies
  • suggest_optimizations
  • create_ticket

The agent operates step-by-step, calling tools to gather data, make decisions, and take real actions.

Challenges we ran into

  • Designing a truly agentic system (not just a chatbot)
  • Balancing autonomy vs control when taking actions
  • Simulating realistic cloud cost scenarios
  • Translating technical signals into business-relevant insights
  • Ensuring reliable tool orchestration and sequencing

Accomplishments that we're proud of

  • Built an agent that closes the loop from insight to action
  • Successfully integrated reasoning + memory + execution
  • Created a realistic FinOps use case with clear ROI

Demonstrated:

  • automated anomaly detection
  • actionable optimization recommendations r- eal ticket creation in GitLab

We moved from “Here’s the problem” to “It’s already being fixed.”

What we learned

  • Agents need tools and structure, not just prompts
  • Memory is key to building compounding intelligence
  • The biggest value comes from automation of decisions and actions
  • Framing outputs in business impact terms is critical
  • Real-world use cases outperform generic AI demos

What's next for Autonomous FinOps Agent

  • Integrate real cloud billing APIs (AWS, GCP, Azure)
  • Add predictive cost forecasting and anomaly prevention
  • Introduce human-in-the-loop approvals for critical actions

Expand into:

  • performance optimization
  • sustainability (carbon-aware FinOps)
  • Build a real-time dashboard for visibility and control

Our vision:

A digital FinOps analyst that continuously monitors, decides, and acts — enabling organizations to control cloud spend at scale.

Built With

Share this project:

Updates