AI-Powered Jenkins Integration with MCP

Tonko Knez

AI Automation

DevOps Engineer

Software Engineer

Docker

Jenkins

Computer Software

mcp-jenkins-adapter

This project demonstrates a fully containerized MCP client-server application that interacts with Jenkins via an API token, and leverages OpenAI for plugging in AI-powered functionalities in Jenkins.

Project Overview

Main feature of the project is using OpenAI through the MCP Client for processing or summarizing Jenkins data.

Jenkins – Main CI/CD server, configured with functional MCP user for automated job operations.

MCP Server – provides AI-powered tools, connects to Jenkins and exposes a SSE endpoint for clients.

MCP Client – Handles interaction with OpenAI and connects to MCP server via SSE, bridging those services together. OpenAI functionality is accessed through MCP Client.

OpenAI – Using provided AI-powered tools to trigger jobs and process Jenkins job data.

The MCP server uses a Jenkins API token generated at Jenkins startup.

Architecture Overview

flowchart TB
    %% Jenkins Container
    subgraph Jenkins_Container["Jenkins Container"]
        JCore["🖥️ Jenkins Core"]
        JPlugins["🔌 Plugins (optional)"]
    end

    %% MCP Server Container
    subgraph MCP_Server_Container["MCP Server Container"]
        MCPServer["🛠️ MCP Server"]
    end

    %% MCP Client Container
    subgraph MCP_Client_Container["MCP Client Container"]
        MCPClient["MCP Client"]
    end

    %% OpenAI External Service
    subgraph OpenAI_Service["OpenAI API (External)"]
        OpenAI["🤖 AI API"]
    end

    %% Connections
    MCPServer -->|Triggers jobs / Queries status| JCore
    JCore -->|Job status / Console log| MCPServer

    MCPClient <-->|SSE / AI Tools| MCPServer
    MCPClient <-->|AI requests / AI Tools| OpenAI

Folder Structure

jenkins-mcp-adapter/
├── docker-compose.yaml
├── Dockerfile.jenkins
├── init.groovy.d
│   ├── create-users.groovy
│   └── seed-job.groovy
├── LICENSE
├── mcp-client
│   ├── Dockerfile
│   ├── go.mod
│   ├── go.sum
│   ├── main.go
├── mcp-server
│   ├── Dockerfile
│   ├── go.mod
│   ├── go.sum
│   └── main.go
└── README.md

Prerequisites

Podman/Docker

Podman/Docker Compose

.env file with required variables (see below)

OpenAI API key

Environment Variables (`.env`)

Create a .env file in the project root with at least the following:

JENKINS_ADMIN_USER=<user>
JENKINS_ADMIN_PASS=<password>
JENKINS_MCP_USER=<mcp_user>
JENKINS_MCP_PASS=<mcp_password>
OPENAI_API_KEY=<your_openai_api_key>

Running the Stack

# First initialize jenkins server to get mcp-user and API key created
$ podman-compose up --build -d jenkins
# Create dir in root of repository and file where token will be stored
$ mkdir -p ./jenkins-secrets
$ touch ./jenkins-secrets/mcp-user.token
# Get API token from jenkins and store it in mcp-user.token file
$ podman-compose exec jenkins cat /var/jenkins_home/secrets/mcp-user.token
# Get rest of services running
$ podman-compose up --build -d

Accessing Services

Jenkins UI: http://localhost:8082/jenkins

MCP Server SSE endpoint: http://localhost:8081/sse

MCP Client:

$ podman-compose exec mcp-client sh # opens terminal in mcp-client container
$ mcp-client # starts client session
MCP initialized. Server: {Name:jenkins-mcp Version:1.0.0}
Jenkins LLM Bridge started. Type your prompts:
> start demo-job # request example forwarded to OpenAI

Processing Jenkins Jobs Data

MCP server exposes tools to trigger Jenkins jobs and process build data:

trigger_job – Trigger a job by name, optionally with parameters.

get_build_status – Get the latest build status for a job.

get_console_log – Retrieve console log for a specific build number.

analyze_logs - A helper function that gathers logs from Jenkins via the MCP Server, transforms them into mcp.Content, and then sends them to OpenAI for analysis.

Role in architecture: Not exposed as a public MCP tool; instead, it’s a private “bridge” feature that enriches what the MCP client can request.

Notes / Best Practices

Always ensure jenkins-secrets/mcp-user.token exists before starting the client-server stack.

.env values are loaded dynamically for security.

Podman Compose handles service dependencies, but Jenkins must initialize before MCP server.

Avoid storing sensitive credentials directly in docker-compose.yml.

Stopping the Stack

$ podman-compose down

This will stop all containers while preserving the volumes (jenkins_home).

Like this project

Posted Sep 14, 2025

Integrated AI functionalities into Jenkins using MCP client-server architecture.

Likes

Views