About jenkins-mcp-enterprise

The most advanced Jenkins MCP server available - Enterprise debugging, multi-instance management, AI-powered failure analysis, vector search, and configurable diagnostics for complex CI/CD pipelines.

j

Published by

jordan-jarvis

Visit View Profile

README.md

View on GitHub

Jenkins MCP Server Enterprise

Jenkins MCP server for multi-instance routing, build diagnostics, and optional vector search.

A Model Context Protocol (MCP) server for Jenkins that provides build management, log inspection, failure diagnostics, and multi-instance configuration. The maintained setup paths in this repository are source install and Docker/Compose.

What It Does

Trigger builds synchronously or asynchronously
Find jobs before triggering or inspecting them
Inspect job definitions and source locations
Retrieve logs and run targeted log-search tools
Discover downstream and sub-build hierarchies
Inspect recent builds or a window around a specific build
Diagnose build failures with configurable recommendations
Route each tool call to the correct Jenkins instance based on jenkins_url
Expose semantic_search only when vector search is enabled
Expose apply_job_edit only when server-side job editing is enabled

Quick Start

Prerequisites

Python 3.10+
Jenkins API access
Jenkins username and API token
Docker and Docker Compose if you want container deployment
Qdrant if you want semantic_search

Install from Source

git clone https://github.com/Jordan-Jarvis/jenkins-mcp-enterprise
cd jenkins-mcp-enterprise
python3 -m pip install -e .

Understand the Two Config Layers

The server uses two configuration layers:

Primary server config: config/mcp-config.yml
Diagnostic config: loaded from one of:
- --diagnostic-config /path/to/file.yml
- JENKINS_MCP_DIAGNOSTIC_CONFIG=/path/to/file.yml
- config/diagnostic-parameters.yml
- bundled defaults in jenkins_mcp_enterprise/diagnostic_config/diagnostic-parameters.yml

You always need the primary server config. You only need a project-local diagnostic config file if you want to override the bundled diagnostic defaults.

Create the Primary Server Config

mkdir -p config
cp config/mcp-config.example.yml config/mcp-config.yml

Edit config/mcp-config.yml with your Jenkins URLs and credentials.

If you want the optional Jenkins-managed job edit workflow for non-SCM-backed jobs, also set:

settings:
  enable_job_editing: true
  job_edit_workspace_dir: "/tmp/mcp-jenkins/job-definitions"

Optional Project-Local Diagnostic Override

Create this only if you want to tune diagnose_build_failure:

cat > config/diagnostic-parameters.yml << 'ENDDIAG'
semantic_search:
  min_diagnostic_score: 0.65
  max_total_highlights: 4

recommendations:
  max_recommendations: 5
ENDDIAG

If you want semantic search, start the local vector stack and set disable_vector_search: false with vector.host: "http://localhost:6333":

./scripts/start_dev_environment.sh

Docker and Compose

cp config/mcp-config.example.yml config/mcp-config.yml
./start-jenkins_mcp_enterprise.sh

See README-Docker.md for the container deployment flow.

Start the Server

jenkins_mcp_enterprise --config config/mcp-config.yml

Connect to Claude Desktop

Add to ~/.claude_desktop_config.json:

{
  "mcpServers": {
    "jenkins": {
      "command": "jenkins_mcp_enterprise",
      "args": ["--config", "config/mcp-config.yml"]
    }
  }
}

Usage Notes

Pass full Jenkins build URLs when a tool or prompt relies on jenkins_url.
In multi-Jenkins setups, each tool call resolves exactly one configured instance from jenkins_url. The server does not fan out across all configured Jenkins instances for a single call.
semantic_search is available only when vector search is enabled.
apply_job_edit is available only when settings.enable_job_editing: true.
get_job_definition returns SCM location details for SCM-backed pipelines. For inline pipelines, it stages a local Groovy file for patch/edit/upload. For other Jenkins-managed jobs, it stages config.xml.
diagnose_build_failure always uses a diagnostic config layer. If you do not provide a project-local override, the bundled defaults are used.
The Docker image includes ripgrep, so ripgrep_search and navigate_log work in container deployments.

Common Usage Patterns

Analyze this failed build: https://jenkins.company.com/job/api-service/456/
Find the root cause in this nested pipeline: https://jenkins.company.com/job/monorepo/job/main/789/
Show me the test failure section from this build: https://jenkins.company.com/job/tests/321/
Find similar authentication failures in recent builds

Available Tools

Tool	Purpose
`trigger_build`	Start a build and wait for completion
`trigger_build_async`	Queue a build without waiting for completion
`trigger_build_with_subs`	Trigger a build and track downstream/sub-build execution
`get_jenkins_job_parameters`	Inspect job parameters before triggering builds
`find_jobs`	Search jobs by name, path, or URL on one resolved Jenkins instance
`list_job_builds`	List recent builds for a job, or a small window around a target build number
`get_build_info`	Fetch metadata for a specific build or `lastBuild`
`get_job_definition`	Inspect whether a job is SCM-backed, inline, multibranch, or XML-backed
`ripgrep_search`	Search logs with regex and context windows
`filter_errors_grep`	Filter logs with common error-oriented patterns
`navigate_log`	Jump to sections or occurrences inside a log
`get_log_context`	Fetch targeted log ranges or chunks
`diagnose_build_failure`	AI-assisted failure diagnosis using logs, hierarchy data, and configured recommendations
`semantic_search`	Vector-backed similarity search across log chunks when vector search is enabled
`apply_job_edit`	Upload a locally edited staged job definition back to Jenkins when job editing is enabled

Configuration Documentation

Development

# Start local supporting services when vector search is enabled
./scripts/start_dev_environment.sh

# Run tests
python3 -m pytest tests/ -v

# Format code
python3 -m black .

License

GPL v3

jenkins-mcp-enterprise