DevOps-OS MCP Server

The DevOps-OS Model Context Protocol (MCP) server turns any MCP-compatible AI assistant into a DevOps automation engine. Ask it in plain English and it generates production-ready CI/CD pipelines, Kubernetes manifests, SRE dashboards, and more — no YAML knowledge required.

Available Tools

Tool	What it generates
`generate_github_actions_workflow`	GitHub Actions CI/CD workflow YAML
`generate_jenkins_pipeline`	Jenkins Declarative Pipeline (Jenkinsfile)
`generate_gitlab_ci_pipeline`	GitLab CI/CD pipeline (`.gitlab-ci.yml`)
`generate_k8s_config`	Kubernetes Deployment + Service manifests
`generate_argocd_config`	Argo CD Application / AppProject or Flux CRs
`generate_sre_configs`	Prometheus alert rules, Grafana dashboards, SLO manifests, Alertmanager routing/config YAML
`scaffold_devcontainer`	`devcontainer.json` + `devcontainer.env.json`
`generate_unittest_config`	Unit test configs for pytest, Jest, Vitest, Mocha, Go

Quick Start — Which Method Is Right for You?

Your situation	Recommended method
Use Claude Code (CLI)	Automated setup script
Use Claude Desktop (GUI app)	Claude Desktop config
Use Cursor IDE	Cursor MCP config
Use VS Code with GitHub Copilot	VS Code MCP config
Use Windsurf	Windsurf config
Use Zed editor	Zed config
Already cloned the repo	Add `--local` to any script below

Step 0 — Prerequisites (all methods)

Requirement	Minimum version	Check
Python	3.10+	`python3 --version`
pip	any recent	`pip --version`
git	any recent	`git --version`

You do not need Docker, Kubernetes, or any cloud accounts.

Option A — Automated Setup (Claude Code CLI)

The fastest path. One command clones the repo, installs dependencies, and registers the server.

# Download the setup script
curl -fsSLo setup_devops_os_mcp.sh https://raw.githubusercontent.com/cloudengine-labs/devops_os/main/mcp_server/setup_devops_os_mcp.sh

# (Optional but recommended) Inspect the script before running it
less setup_devops_os_mcp.sh

# Run the setup script
bash setup_devops_os_mcp.sh

Already cloned the repo? Run the script in local mode instead of downloading it:

# From inside your devops_os clone:
bash mcp_server/setup_devops_os_mcp.sh --local

Custom install directory:

INSTALL_DIR=~/projects/devops_os bash mcp_server/setup_devops_os_mcp.sh

Verify it worked:

claude mcp list          # should show "devops-os"
claude mcp get devops-os # shows the full config

Test it:

claude
> Generate a GitHub Actions workflow for a Python + Node.js app with Docker build and Kubernetes deployment

Option B — Claude Desktop App

Clone and install:

git clone https://github.com/cloudengine-labs/devops_os.git ~/devops_os
cd ~/devops_os
python3 -m venv .venv
source .venv/bin/activate          # macOS/Linux
# .venv\Scripts\activate           # Windows
pip install -r mcp_server/requirements.txt

Find your config file:

OS Path

macOS ~/Library/Application Support/Claude/claude_desktop_config.json

Windows %APPDATA%\Claude\claude_desktop_config.json

Linux ~/.config/Claude/claude_desktop_config.json

OS	Path
macOS	`~/Library/Application Support/Claude/claude_desktop_config.json`
Windows	`%APPDATA%\Claude\claude_desktop_config.json`
Linux	`~/.config/Claude/claude_desktop_config.json`

Add this block (replace /YOUR/HOME with the actual path from step 1):

{
  "mcpServers": {
    "devops-os": {
      "command": "/YOUR/HOME/devops_os/.venv/bin/python",
      "args": ["-m", "mcp_server.server"],
      "cwd": "/YOUR/HOME/devops_os"
    }
  }
}

Tip: Use the full venv path, not just python, so the correct dependencies are used.

Restart Claude Desktop, then ask:

"Generate a Jenkins pipeline for a Java Spring Boot app that builds a Docker image and deploys to Kubernetes."

Option C — Cursor IDE

Clone and install (same as step 1 above).

Create .cursor/mcp.json in your project root (or copy from the devops_os repo):

{
  "mcpServers": {
    "devops-os": {
      "command": "/YOUR/HOME/devops_os/.venv/bin/python",
      "args": ["-m", "mcp_server.server"],
      "cwd": "/YOUR/HOME/devops_os"
    }
  }
}

Reload Cursor (Cmd+Shift+P → "Developer: Reload Window").
Open the Cursor chat panel and ask:

"Scaffold a devcontainer for a Go + Python project with Terraform and kubectl."

Project-level config: The .cursor/mcp.json in this repo is pre-configured and works if you clone devops_os and open it directly in Cursor.

Option D — VS Code + GitHub Copilot

VS Code 1.99+ supports MCP servers in GitHub Copilot Agent mode.

Clone and install (same as step 1 in Option B).

Create .vscode/mcp.json in your workspace:

{
  "servers": {
    "devops-os": {
      "type": "stdio",
      "command": "/YOUR/HOME/devops_os/.venv/bin/python",
      "args": ["-m", "mcp_server.server"],
      "cwd": "/YOUR/HOME/devops_os"
    }
  }
}

Open GitHub Copilot Chat in Agent mode and ask:

"Use devops-os to generate Kubernetes manifests for my app."

Note: VS Code MCP support requires GitHub Copilot and VS Code 1.99 or later.

Option E — Windsurf

Clone and install (same as step 1 in Option B).
Open Windsurf settings: Windsurf → Preferences → Cascade → MCP Servers.

Add this configuration:

{
  "mcpServers": {
    "devops-os": {
      "command": "/YOUR/HOME/devops_os/.venv/bin/python",
      "args": ["-m", "mcp_server.server"],
      "cwd": "/YOUR/HOME/devops_os"
    }
  }
}

Alternatively, edit ~/.codeium/windsurf/mcp_config.json directly.

Restart Windsurf and ask Cascade:

"Generate SRE configs for my payment-service with a 99.9% SLO target."

Option F — Zed Editor

Clone and install (same as step 1 in Option B).

Open ~/.config/zed/settings.json and add:

{
  "context_servers": {
    "devops-os": {
      "command": {
        "path": "/YOUR/HOME/devops_os/.venv/bin/python",
        "args": ["-m", "mcp_server.server"],
        "env": {
          "PYTHONPATH": "/YOUR/HOME/devops_os"
        }
      }
    }
  }
}

Restart Zed and use the Assistant panel to ask:

"Create a GitLab CI pipeline for a Python microservice."

Example Prompts

Once connected, try these prompts in any AI tool:

Generate a complete GitHub Actions CI/CD workflow for a Python + Node.js project
with Docker build and Kubernetes deployment using Kustomize.

Create a Jenkins pipeline for a Java Spring Boot service with Maven build,
Docker image push to ECR, and ArgoCD deployment.

Generate Kubernetes manifests for 'api-service' using image
'ghcr.io/myorg/api-service:v1.2.3' with 3 replicas on port 8080.

Scaffold a devcontainer for a Go + Python project with Terraform, kubectl, and k9s.

Generate SRE configs (Prometheus alerts, Grafana dashboard, SLO manifest)
for 'payment-service' owned by team 'platform' with 99.9% availability target.

Create unit test configuration for a Python project named 'data-pipeline'.

Generate a GitLab CI pipeline with Docker build and Kubernetes deployment stages
for a Python microservice.

Architecture

AI Assistant (Claude / Cursor / VS Code / Windsurf / Zed)
        │  MCP stdio request
        ▼
DevOps-OS MCP Server  (mcp_server/server.py)
        │  calls Python scaffold functions
        ▼
DevOps-OS CLI modules
  ├─ cli/scaffold_gha.py         → GitHub Actions YAML
  ├─ cli/scaffold_jenkins.py     → Jenkinsfile
  ├─ cli/scaffold_gitlab.py      → .gitlab-ci.yml
  ├─ cli/scaffold_argocd.py      → ArgoCD / Flux manifests
  ├─ cli/scaffold_sre.py         → Prometheus / Grafana / SLO
  ├─ cli/scaffold_devcontainer.py→ devcontainer.json
  ├─ cli/scaffold_unittest.py    → pytest / Jest / Go test
  └─ kubernetes/k8s-config-generator.py → K8s manifests

Troubleshooting

"Module not found" or "No module named mcp"

The venv is missing or incomplete. Re-run:

/path/to/devops_os/.venv/bin/pip install -r /path/to/devops_os/mcp_server/requirements.txt

"command not found: claude" (during setup)

Install Claude Code first: claude.ai/code

Server shows in `claude mcp list` but tools don't appear

Check that PYTHONPATH points to the repo root:

claude mcp get devops-os   # inspect the registered command

Re-register manually if needed:

claude mcp remove devops-os
claude mcp add --transport stdio devops-os --scope user \
  -- /path/to/devops_os/.venv/bin/python -m mcp_server.server \
  --env PYTHONPATH=/path/to/devops_os

Tools appear but return errors

Run the test suite to verify the installation:

cd /path/to/devops_os
.venv/bin/python -m pytest mcp_server/test_server.py -v

Windows path format

On Windows, use forward slashes or escaped backslashes in JSON config:

"command": "C:/Users/you/devops_os/.venv/Scripts/python.exe"

Using with Claude API / OpenAI function calling

See ../skills/README.md for examples of calling the DevOps-OS tools directly from the Anthropic or OpenAI Python SDKs (no MCP client required).

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

DevOps-OS MCP Server

Available Tools

Quick Start — Which Method Is Right for You?

Step 0 — Prerequisites (all methods)

Option A — Automated Setup (Claude Code CLI)

Option B — Claude Desktop App

Option C — Cursor IDE

Option D — VS Code + GitHub Copilot

Option E — Windsurf

Option F — Zed Editor

Example Prompts

Architecture

Troubleshooting

"Module not found" or "No module named mcp"

"command not found: claude" (during setup)

Server shows in `claude mcp list` but tools don't appear

Tools appear but return errors

Windows path format

Using with Claude API / OpenAI function calling

FilesExpand file tree

README.md

Latest commit

History

README.md

File metadata and controls

DevOps-OS MCP Server

Available Tools

Quick Start — Which Method Is Right for You?

Step 0 — Prerequisites (all methods)

Option A — Automated Setup (Claude Code CLI)

Option B — Claude Desktop App

Option C — Cursor IDE

Option D — VS Code + GitHub Copilot

Option E — Windsurf

Option F — Zed Editor

Example Prompts

Architecture

Troubleshooting

"Module not found" or "No module named mcp"

"command not found: claude" (during setup)

Server shows in claude mcp list but tools don't appear

Tools appear but return errors

Windows path format

Using with Claude API / OpenAI function calling

Server shows in `claude mcp list` but tools don't appear