DevOps MCP Server

A FastMCP-based MCP server providing DevOps tools and integrations.

This a conservative MCP server. It does not add, update or delete anything in your system, does not run any job. Basically, it is read-only. It only retrieves data for analysis, display the information.

So it is safe for DevOps.

Features

GitHub Integration

• Repository search and management
• File content retrieval from repositories
• Issue tracking and management
• Code search functionality
• Commit history viewing

Jenkins Integration

• Job listing and management
• Build log retrieval
• View management
• Build parameter inspection
• Recent failed builds monitoring

Installation

To install the package, use the following command:

pip install devops-mcps

Usage

Run the MCP server:

devops-mcps

Configuration

Environment Variables

GitHub Configuration

Set the required environment variable for GitHub API access:

export GITHUB_PERSONAL_ACCESS_TOKEN=your_token_here

Jenkins Configuration

Set the required environment variables for Jenkins API access:

export JENKINS_URL=your_jenkins_url
export JENKINS_USER=your_jenkins_username
export JENKINS_TOKEN=your_jenkins_token
export LOG_LENGTH=5120

Note: LOG_LENGTH means it will retrieve this length of jenkins log for analysis. It does not always the longer the better.

UVX Configuration

Install UVX tools:

uvx install

Run with UVX:

uvx devops-mcps

Transport Configuration

The MCP server supports two transport types:

• stdio (default): Standard input/output communication
• sse: Server-Sent Events for HTTP-based communication

Local Usage

# Default stdio transport
devops-mcps

# SSE transport
devops-mcps --transport sse

UVX Usage

# Default stdio transport
uvx run devops-mcps

# SSE transport
uvx run devops-mcps-sse

Docker Configuration

Build the Docker image:

docker build -t devops-mcps .

Run the container:

docker run -p 8000:8000 devops-mcps

GitHub Public and Enterprise Support

This project supports both public GitHub and GitHub Enterprise automatically.

• By default, it connects to public GitHub (https://api.github.com).
• To use with GitHub Enterprise, set the GITHUB_API_URL environment variable to your enterprise API endpoint (e.g., https://github.mycompany.com/api/v3).

Example:

# For public GitHub (default)
export GITHUB_PERSONAL_ACCESS_TOKEN=your_token_here

# For GitHub Enterprise
export GITHUB_PERSONAL_ACCESS_TOKEN=your_token_here
export GITHUB_API_URL=https://github.mycompany.com

The server will detect the correct API endpoint at runtime.

VSCode Configuration

To use this MCP server in vs code copilot, there are 2 ways to configure it in VSCode settings.json with different transport types:

UVX Configuration

stdio Transport (default)

"devops-mcps": {
  "type": "stdio",
  "command": "uvx",
  "args": ["devops-mcps"],
  "env": {
    "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxxxxxxxxxxxxxxxxxxxxxxxxCe",
    "GITHUB_API_URL": "https://github.mycompany.com",
    "JENKINS_URL": "jenkins_url_here",
    "JENKINS_USER": "jenkins_username_here",
    "JENKINS_TOKEN": "jenkins_password_here"
  }
}

SSE Transport

"devops-mcps": {
  "type": "sse",
  "command": "uvx",
  "args": ["devops-mcps-sse"],
  "env": {
    "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxxxxxxxxxxxxxxxxxxxxxxxxCe",
    "GITHUB_API_URL": "https://github.mycompany.com",
    "JENKINS_URL": "jenkins_url_here",
    "JENKINS_USER": "jenkins_username_here",
    "JENKINS_TOKEN": "jenkins_password_here"
  }
}

Docker Configuration

stdio Transport (default)

"devops-mcps": {
  "command": "docker",
  "args": [
    "run",
    "-i",
    "huangjien/devops-mcps:latest"
  ],
  "env": {
    "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxxxxxxxxxxxxxxxxxxxxxxxx2Ce",
    "GITHUB_API_URL": "https://github.mycompany.com",
    "JENKINS_URL": "jenkins_url_here",
    "JENKINS_USER": "jenkins_username_here",
    "JENKINS_TOKEN": "jenkins_password_here"
  }
}

SSE Transport (MCP Server Deployed in Remote Docker Container)

"devops-mcps": {
  "type": "sse",
  "url": "http://[remote ip address]:8000/sse",
  "env": {
    "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxxxxxxxxxxxxxxxxxxxxxxxx2Ce",
    "GITHUB_API_URL": "https://github.mycompany.com",
    "JENKINS_URL": "jenkins_url_here",
    "JENKINS_USER": "jenkins_username_here",
    "JENKINS_TOKEN": "jenkins_password_here"
  }
}

Note: The docker should start like:

docker run -p 8000:8000 -e TRANSPORT_TYPE=sse -i huangjien/devops-mcps:latest

Development

Install development dependencies:

uv pip install -e .[dev]

or

uv sync

Recommend to install vs code extension: ruff

Or do it in command line:

To lint (check):

uvx ruff check

To format:

uvx ruff format

Run mcp inspector to test or debug:

npx @modelcontextprotocol/inspector uv run devops-mcps

CI/CD Pipeline

GitHub Actions workflow will automatically:

1. Build and publish Python package to PyPI
2. Build and push Docker image to Docker Hub

Required Secrets

Set these secrets in your GitHub repository:

• PYPI_API_TOKEN: Your PyPI API token
• DOCKER_HUB_USERNAME: Your Docker Hub username
• DOCKER_HUB_TOKEN: Your Docker Hub access token

Workflow triggers on push to main branch.

Packaging and Publishing

Install tools

pip install -U twine build  

Build the package

python -m build

Upload to PyPI

1. Create a ~/.pypirc file with your API token:

   [pypi]
   username = __token__
   password = your_pypi_api_token_here
   

2. Upload the package:

   twine upload dist/*
   

Important Notes

• Ensure all classifiers in pyproject.toml are valid PyPI classifiers
• Remove deprecated license classifiers in favor of SPDX license expressions
• The package will be available at: https://pypi.org/project/devops-mcps/
• Update the version everytime, or when you push, it will show an error: already exists.

License

MIT

Appendix (Query in Github)

For Repository Search (gh_search_repositories):

• in:: Search only in specific fields (name, description, readme, or combinations). query="fastapi in:name" (Find repos with "fastapi" in their name) query="web framework in:readme,description"

• user: or org:: Search within a specific user's or organization's repositories. query="user:tiangolo fastapi"

• language:: Filter by programming language. query="http client language:python"

• stars:<N>, forks:<N>: Filter by number of stars or forks (use ranges like >100, <50, 10..50). query="language:javascript stars:>1000"

• created:<YYYY-MM-DD>, pushed:<YYYY-MM-DD>: Filter by creation or last push date (use ranges). query="data science pushed:>2024-03-01"

• topic:: Filter by repository topic. query="topic:docker topic:python"

• license:: Filter by license (e.g., mit, apache-2.0). query="language:go license:mit"

For Code Search (gh_search_code):

• in:: Search within file content (default), path, or both. q='"import requests" in:file'

• user: or org:: Search code within a specific user's or organization's repositories. q='"BaseSettings" user:tiangolo'

• repo:/: Search code within a specific repository. q='"JenkinsAPIException" repo:devops-mcps/devops-mcps'

• language:: Filter by the language of the file containing the code. q='def main language:python'

• path:, path:/: Search within specific paths or the root directory. q='"GithubException" path:src/devops_mcps'

• filename:: Search within specific filenames. q='TODO filename:core.py'

• extension:: Search within files having a specific extension. q='class Settings extension:py'

This query syntax provides a flexible and powerful way to find exactly what you need on GitHub directly through your MCP tools. You can find the full, official documentation here:

• Searching on GitHub: https://docs.github.com/en/search-github/searching-on-github
• Searching Code: https://docs.github.com/en/search-github/searching-on-github/searching-code
• Searching Repositories: https://docs.github.com/en/search-github/searching-on-github/searching-for-repositories