AI Integration — MCP Server & AI Skills#

DevOps-OS exposes all its pipeline automation tools as an MCP (Model Context Protocol) server and as AI skill definitions for Claude and OpenAI.

Available Tools#

MCP Server#

Installation#

pip install -r mcp_server/requirements.txt

Running the server#

# Run as a stdio MCP server (default — for Claude Desktop and most MCP clients)
python -m mcp_server.server

# Or directly
python mcp_server/server.py

Connecting to Claude Desktop#

Add to claude_desktop_config.json
(~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "devops-os": {
      "command": "python",
      "args": ["-m", "mcp_server.server"],
      "cwd": "/path/to/devops_os"
    }
  }
}

Restart Claude Desktop, then ask it:

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

“Create a Jenkins pipeline for a Python microservice with Docker build and push stages.”

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

Architecture#

AI Assistant (Claude / ChatGPT)
        │  MCP / function-call request
        ▼
DevOps-OS MCP Server  ←──────────┐
        │                        │
        │  calls Python functions │
        ▼                        │
  devopsos scaffold gha          │
  devopsos scaffold gitlab  (same generators
  devopsos scaffold jenkins  used by the CLI)
  devopsos scaffold argocd        │
  devopsos scaffold sre           │
        │                        │
        ▼                        │
  Generated files ───────────────┘

Using with Claude (Anthropic API)#

Load skills/claude_tools.json as the tools parameter:

import json
import anthropic

with open("skills/claude_tools.json") as fh:
    tools = json.load(fh)

client = anthropic.Anthropic()
response = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=4096,
    tools=tools,
    messages=[{
        "role": "user",
        "content": (
            "Generate a complete GitHub Actions CI/CD workflow for a "
            "Python + Node.js project with Kubernetes deployment via Kustomize."
        ),
    }],
)

for block in response.content:
    if block.type == "tool_use":
        print(f"Tool: {block.name}")
        print(f"Input: {json.dumps(block.input, indent=2)}")

Using with OpenAI (function calling)#

Load skills/openai_functions.json as the tools parameter:

import json
import openai

with open("skills/openai_functions.json") as fh:
    functions = json.load(fh)

client = openai.OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    tools=functions,
    messages=[{
        "role": "user",
        "content": "Generate a Jenkins pipeline for a Java Spring Boot app."
    }],
)

for choice in response.choices:
    if choice.message.tool_calls:
        for tc in choice.message.tool_calls:
            print(f"Function: {tc.function.name}")
            print(f"Args: {tc.function.arguments}")

Custom GPT / GPT Actions#

Use skills/openai_functions.json as the OpenAPI schema for a Custom GPT Action:

1. Open ChatGPT → Create a GPT → Configure → Actions → Create new action
2. Paste the contents of skills/openai_functions.json
3. Set the server URL to your deployed MCP server endpoint
4. Save and test the GPT

Example Prompts#

Generate a GitHub Actions workflow for a Java Spring Boot app with kubectl deployment.

Create a Jenkins pipeline for a Python microservice with Docker build and push stages.

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

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

Create a GitLab CI pipeline with Python testing and ArgoCD deployment.

Generate SRE configs for my-service with a 99.9% availability SLO and PagerDuty alerts.