← Agentic AI for DevOps

Build what real companies run

Every project clones a system that's actually in production at a named company — researched and verified. You build the real architecture, not a toy demo.

Season 1 · 10 projects · grounded in Datadog, Resolve AI, Uber, Block, CodeRabbit, GitHub, Pulumi & CAST AI

Tier 1Flagship Anchor

Highest production credibility — start here. These set the architecture everything else reuses.

🚨
Building now

AI SRE — Autonomous Incident Investigator

The agent that finds root cause before you finish your coffee

An AI SRE that watches your alerts, then investigates across metrics, logs, and Kubernetes in an agentic tool-use loop — forming and testing hypotheses until it posts a ranked root-cause analysis with the evidence trail and a suggested fix to Slack. Read-only by design, so a human approves any action.

MirrorsDatadog Bits AI SREResolve AICleric AI

✓ Resolve AI is a $1B unicorn used by Coinbase & DoorDash; Datadog Bits AI SRE is GA across thousands of orgs.

Build on: The open-source HolmesGPT pattern (Apache 2.0, CNCF Sandbox, by Robusta + Microsoft).

Claude Tool UsePython / FastAPIPrometheusLokiKubernetesSlack
Intermediate⏱ ~3 hrs
View build guide →
🔌
On the roadmap

MCP DevOps Agent — your own “goose for ops”

One agent, every tool — via Model Context Protocol

An MCP-native agent that connects to your real infrastructure through MCP servers (kubectl, AWS, GitHub, Prometheus) and executes ops tasks end-to-end with human approval. The exact pattern Block and Uber run internally at massive scale.

MirrorsBlock “codename goose”Uber MCP platform

✓ Block's goose now writes ~15% of all Block production code; Uber runs 60,000+ agent tasks/week on MCP.

Build on: Block's open-source goose (Apache 2.0) + official MCP servers.

ClaudeModel Context Protocolkubectl MCPAWS MCPGitHub MCP
Intermediate⏱ ~3 hrs
Notify me ↗
📚
On the roadmap

RAG On-Call Copilot (Slack)

Answers on-call questions from your runbooks, instantly

A retrieval-augmented copilot that ingests your runbooks, wikis, and past incidents, then answers engineers' on-call questions in Slack with cited sources — cutting the “who knows about X?” tax. Uber's version saved ~13,000 engineering hours.

MirrorsUber GenieMoveworks

✓ Uber's Genie answered 70,000+ questions across 154 Slack channels and saved ~13,000 engineering hours.

Build on: Standard RAG stack — your docs + embeddings + a vector DB + Claude.

ClaudeEmbeddingsVector DB (pgvector/Qdrant)SlackRAG
Beginner⏱ ~2 hrs
Notify me ↗

Tier 2Code & Pipeline

Dev-facing agents with the biggest audience overlap — PR review, security, self-healing CI/CD.

🔍
On the roadmap

Agentic PR Reviewer

An AI reviewer that actually reads your codebase

A GitHub bot that reviews pull requests like a senior engineer — running shell and Python in a sandbox to navigate the diff, trace symbols, and leave inline comments. Not a single prompt: a real code-execution agent.

MirrorsCodeRabbitQodoGreptile

✓ CodeRabbit runs this on Google Cloud Run at 200+ instances and 10 req/s peak in production.

Build on: GitHub App + a sandboxed code-execution loop (cat/grep/ast-grep).

Claude Tool UseGitHub APISandbox (Docker/microVM)Node.js / Python
Advanced⏱ ~3 hrs
Notify me ↗
🛡️
On the roadmap

AI Security Autofix

Find a vulnerability, ship the fix PR automatically

An agent that runs a SAST scanner (Semgrep/CodeQL), feeds each finding plus the surrounding code-flow to an LLM, and generates a verified fix as a pull request — the architecture GitHub ships to millions of repos.

MirrorsGitHub Copilot AutofixWizSnyk AI

✓ GitHub Copilot Autofix (CodeQL + GPT-4o) cut median remediation time ~3x in production.

Build on: Open SAST (Semgrep) → LLM fix → PR, wired into CI.

ClaudeSemgrep / CodeQLGitHub ActionsPython
Intermediate⏱ ~2.5 hrs
Notify me ↗
🔁
On the roadmap

Self-Healing CI/CD Pipeline

A pipeline that fixes its own broken builds

When a GitHub Actions run fails, an agent reads the logs, reproduces the error, writes a fix, and opens a pull request — closing the loop on flaky builds and trivial breakages, with a human merging.

MirrorsGitHub Copilot coding agentself-healing DevOps pattern

✓ A trending, validated format — freeCodeCamp & top DevOps creators shipped this exact build in mid-2026.

Build on: GitHub Actions failure webhook → agent → fix PR.

Claude APIGitHub ActionsNode.jsOctokit
Advanced⏱ ~3 hrs
Notify me ↗

Tier 3Infra & Cost

Platform-engineering track — Kubernetes, Terraform, and cloud cost agents.

☸️
On the roadmap

Kubernetes Copilot

Talk to your cluster in plain English

A tool-use agent that diagnoses CrashLoopBackOff and OOMKilled pods, translates natural language into safe kubectl operations, and explains what went wrong with real cluster context — the k8sgpt idea, agentic.

Mirrorsk8sgpt (CNCF)Komodor Klaudiakubectl-ai

✓ k8sgpt is a CNCF project with 6k+ stars; the NL-to-cluster-diagnosis pattern is widely adopted.

Build on: k8sgpt-style analyzers + a Claude tool-use loop over the K8s API.

Claude Tool UseKubernetes APIPythonHelm
Intermediate⏱ ~2.5 hrs
Notify me ↗
🏗️
On the roadmap

Terraform Review & Drift Agent

An AI senior platform engineer reviews your IaC

An agent that reviews `terraform plan` on every PR — flagging security misconfigurations, cost spikes, and drift — and drafts safe fixes, gated behind policy guardrails and human approval. The Pulumi Neo / Firefly pattern.

MirrorsPulumi NeoFirefly

✓ Pulumi Neo and Firefly ship production agents with MCP servers and human-in-the-loop guardrails.

Build on: terraform plan (JSON) → LLM review → inline PR comments + drafted fix.

ClaudeTerraform / OpenTofuGitHubOPA / policy
Intermediate⏱ ~2.5 hrs
Notify me ↗
💸
On the roadmap

AI Cloud Cost / FinOps Agent

Turn AWS bills into automated savings

An agent that ingests AWS Cost & Usage reports plus Kubernetes utilization, finds idle and over-provisioned resources, estimates monthly savings, and drafts the Terraform/CLI changes to act on them.

MirrorsCAST AIVantageAntimetal

✓ CAST AI autonomously cuts Kubernetes cost 40–70% in production (e.g. Akamai).

Build on: Cost & usage data + OpenCost → LLM analysis → savings report + fix drafts.

ClaudeAWS Cost ExplorerKubecost / OpenCostPython / boto3
Advanced⏱ ~3 hrs
Notify me ↗

Tier 4Capstone

The season finale — orchestrate every specialist agent into one autonomous AI DevOps team.

🧠
On the roadmap

Multi-Agent Ops Platform

An autonomous AI DevOps team — the season finale

A planner agent that orchestrates every specialist you built (SRE, K8s, IaC, FinOps, security) into one coordinated system — routing an incident to the right agents, merging their findings, and proposing a plan. The Resolve AI / AWS Bedrock AgentCore architecture.

MirrorsResolve AIAWS Bedrock AgentCore (multi-agent SRE)

✓ Resolve AI's multi-agent planner-orchestrates-specialists design is the blueprint for autonomous ops.

Build on: A planner/orchestrator over the specialist agents from projects 1–9.

ClaudeMulti-agent orchestrationMCPPlanner + specialists
Advanced⏱ ~4 hrs
Notify me ↗

Want these the moment they drop?

Subscribe and get each production-grade build — code, video, and write-up — as it ships.

Subscribe to Newsletter

Get the latest articles and tutorials delivered to your inbox.

We respect your privacy. Unsubscribe at any time.