Supercharging DevOps Productivity with Claude/Cursor — MCPs , Plugins & Skills

Ravish Khullar

3 min read

·

Sep 19, 2026

--

1

Listen

Share

Most DevOps engineers already use Claude, Cursor or other AI coding assistants. But using AI only to generate Terraform, fix YAML or write shell scripts barely scratches the surface.

The bigger opportunity is to connect AI with the tools we already use every day:

GitLab → Jenkins → Terraform → AWS → Argo CD → Kubernetes → Grafana → Confluence

With the right MCPs, plugins and skills, Claude or Cursor can become an AI-powered DevOps cockpit rather than just another coding assistant.

The Stack I’m Exploring

I group the tools into a few logical layers.

Press enter or click to view image in full size

The goal isn’t to install every MCP available.

The goal is to connect the engineering lifecycle.

Context Is Everything

One of the biggest AI productivity problems is bad context.

Tools such as Context7, Obsidian and Confluence can help provide documentation and organizational knowledge.

Graphify and Code Review Graph can provide structural understanding of repositories and dependencies.

Similarly, RTK (Rust Token Killer) can reduce noisy terminal output before it enters the model context.

Instead of throwing thousands of lines from:
- kubectl
- terraform
- git
- docker
- grep

into Claude, we send the useful signal.

Better context often matters as much as a better model.

Connect the Complete Delivery Pipeline

The interesting workflow starts when the integrations work together.

Imagine asking: “Investigate why the latest deployment increased 5xx errors”

The agent could:

GitLab
   ↓
Analyze latest changes
   ↓
Code Review Graph
   ↓
Check Jenkins pipeline
   ↓
Check Argo CD deployment
   ↓
Query Grafana
   ↓
Compare before/after metrics
   ↓
Identify probable cause
   ↓
Generate remediation

That is significantly more useful than manually moving between six different tools.

Add Infrastructure and FinOps

Now add Terraform + AWS Billing.

The AI can potentially correlate:

AWS Cost Increase
       ↓
Terraform Change
       ↓
Infrastructure Change
       ↓
Kubernetes Scaling
       ↓
Workload / Deployment

Instead of simply asking: “Why did AWS cost increase?”

we can ask: “What infrastructure changes contributed to the increase, which workloads caused them, and what optimizations are available?”

This brings FinOps directly into the DevOps workflow.

Architecture and Documentation Should Be Automatic

Tools like Excalidraw and draw.io can turn architecture understanding into diagrams.

So after investigating a system:

Analyze Architecture
       ↓
Generate Diagram
       ↓
Generate Runbook
       ↓
Update Confluence
       ↓
Generate PPTX

Remotion can extend the same information into programmatic videos or demos.

One investigation can therefore produce:

code + fix + diagram + documentation + presentation

instead of requiring separate manual work.

Don’t Give AI Unlimited Production Access

Connecting AI to infrastructure doesn’t mean giving it unrestricted control.

I prefer a progressive model:

Read Only
   ↓
Generate Commands
   ↓
Generate GitLab MR
   ↓
Human Approval
   ↓
Controlled Execution
   ↓
Automatic Verification

For production, human approval should remain an important control point, especially for high-impact actions.

AI should remove repetitive work — not remove engineering governance.

My Ideal AI DevOps Workflow

A future prompt could simply be: Investigate the deployment issue.

And behind that request:

Claude / Cursor
      ↓
Context7 + Confluence
      ↓
GitLab + Code Graph
      ↓
Jenkins
      ↓
Terraform + AWS
      ↓
Argo CD + Kubernetes
      ↓
Grafana
      ↓
Root Cause
      ↓
Fix
      ↓
Validation
      ↓
Documentation

The engineer moves from operating tools to orchestrating engineering workflows.

Final Thought

The goal isn’t: “How many MCPs can I install?”

The better question is: “How many repetitive steps between identifying a problem and delivering a verified solution can I eliminate?”

With tools such as —

Ponytail, OmniRoute, Graphify, RTK, UI/UX Pro Max, Superpower, Caveman, Obsidian, AgentFlow, Context7, GEO, AI Crawler, Code Review Graph, GitLab, Jenkins, Terraform, Argo CD, Grafana, AWS Billing, Excalidraw, draw.io, PPTX, Remotion and Confluence.

Claude/Cursor can become much more than a coding assistant. It can become an AI interface for the entire DevOps lifecycle.

Context → Code → Build → Infrastructure → Deploy → Observe → Fix → Document