Jenkins MCP Server

mcp_jenkins

A Jenkins MCP server. Model Context Protocol (MCP) lets AI tools (like chatbots) talk to and control your Jenkins setup, i. e. retrieve information and modify settings.

Note: This is a minimal experimental version of the MCP Jenkins server and is currently in early development.

Description

This project provides a Model Context Protocol (MCP) server for interacting with Jenkins. It allows users to trigger Jenkins jobs, get build statuses, and perform other Jenkins-related operations through the MCP interface.

Components

• server.py: The core MCP Jenkins server application.
• functions_schema.md: Defines the schema for the functions exposed by the MCP Jenkins server.
• client.py: An example client demonstrating how to interact with the MCP Jenkins server (provided for reference only).
• functional tests: Contains functional tests for the MCP Jenkins server.

Installation

To install the package, run:

Usage

Common Workflows

Running the Server

To run the MCP server:

Running the Example Client

To run the example client:

For example, to list builds for a job named "backups" using a specific model, you can run:

Note: If the package is installed via pip install ., the mcp_jenkins_client console script is also available.

This might produce output similar to:

Building and Testing with Docker

A common workflow for development and testing is to first build the Docker image and then execute the tests:

1. Build the Docker image: This step prepares the environment needed for testing.
   ./docker/build
2. Run tests: After the build is complete, execute the tests.
   ./docker/run.tests

This sequence ensures that tests are performed against the latest build in a consistent Dockerized environment.

Deploying a Test Environment

To deploy a local Jenkins testing instance (without authentication, for functional tests only):

Connecting to an Existing Jenkins Instance

To use the MCP Jenkins server with an existing Jenkins instance, you need to configure the following environment variables:

• JENKINS_URL: The full URL of your Jenkins instance (e.g., http://your-jenkins-host:8080). This is required.
• JENKINS_USER: (Optional) Your Jenkins username if authentication is required.
• JENKINS_API_TOKEN: (Optional) Your Jenkins API token. This must be provided along with JENKINS_USER if authentication is used. You can generate an API token in your Jenkins user's configuration page (<Jenkins URL>/me/configure).
• MCP_API_KEY: A secret API key to secure this MCP server. Requests to the MCP server will need to include this key in the X-API-Key header. This is required unless DEBUG_MODE is set to true.
• DEBUG_MODE: Set to true to run the MCP server in debug mode, which bypasses the MCP_API_KEY requirement and provides more verbose logging. Do not use in production.

Example Configuration (Bash):

Once these environment variables are set, you can run the MCP server using the Docker script:

The MCP server will then attempt to connect to your specified Jenkins instance.

OpenWebUI Integration

The file open-webui/open_webui_interface.py provides an example of how to integrate this MCP Jenkins server with an OpenWebUI instance.

To use it:

1. In your OpenWebUI interface, navigate to the section for adding or configuring tools.
2. Create a new tool.
3. Copy the entire content of the open-webui/open_webui_interface.py file and paste it into the tool configuration in OpenWebUI.
4. Important: You will need to adjust the connection parameters within the pasted code, specifically:
   • MCP_JENKINS_SERVER_URL: Set this environment variable in your OpenWebUI environment to the URL of your running MCP Jenkins server (e.g., http://localhost:5000). The script defaults to http://localhost:5000 if the variable is not set.
   • MCP_API_KEY: If your MCP Jenkins server is configured to require an API key, ensure this environment variable is set in your OpenWebUI environment. The script will print a warning if it's not found but will still attempt to make requests.

Once configured, the tools defined in open_webui_interface.py (e.g., list_jobs, trigger_build, get_build_status) should become available for use within your OpenWebUI chat interface.

License

This project is licensed under the MIT License.