Labsco
MCP SERVER

MCP Jenkins Intelligence

by heniv96 ↗

Ask questions about Jenkins pipelines in natural language — health, failures, trends and dependencies — with the sensitive data anonymised before it is analysed.

Build Systems & CI/CD
Summary
The anonymisation is what makes it usable on a real estate.

CI metadata is more sensitive than it looks — pipeline and repository names describe a company's internal structure, and log lines carry tokens. Hashing that before analysis is what lets you point this at a production Jenkins rather than a sandbox. The queue monitor and the dependency view are the tools that pay for themselves fastest when a build is stuck.

What it is

An MCP server for Jenkins that goes past listing builds. Alongside the read tools it carries analysis: pipeline health over a period, failure diagnosis for a specific build, trend comparison across pipelines, anomaly detection and optimisation suggestions. Processing happens locally, with pipeline names, branches, repositories and other identifying data replaced by hashes before analysis.

What you get
  • Read the pipelines: `list_pipelines` with an optional `search`, `get_pipeline_details` and `get_pipeline_builds` filtered by `limit` and `status`
  • Diagnose: `analyze_pipeline_health` over a period, `analyze_pipeline_failure` for one `build_number`, and `ask_pipeline_question` for a natural-language question across named pipelines
  • Monitor: `get_pipeline_metrics`, `get_pipeline_dependencies`, `monitor_pipeline_queue` for the Jenkins build queue and `analyze_build_trends` across several pipelines over a period
  • Look ahead: `predict_pipeline_failure`, `detect_pipeline_anomalies` with an adjustable `sensitivity`, `suggest_pipeline_optimization` and `generate_ai_insights` by insight type
  • Anonymisation that covers pipeline, cluster, folder, application, branch, organisation and repository names, along with build numbers, timestamps, URLs, user information, token values and log entries — applied recursively through nested structures
Requirements

A Jenkins server with API access and a token: `JENKINS_URL`, `JENKINS_USERNAME` and `JENKINS_TOKEN` are all required. No Python installation is needed for the released binaries — download the macOS arm64 or Linux amd64 build, make it executable, and point your client at its path with the three variables in the `env` block. Python 3.8 or higher is only needed for the development setup. All processing is local, with no data sent to external AI services.

Setup effort

One command plus a key — curl -fsSL https://raw.githubusercontent.com/heniv96/mcp-jenkins-intelligence/main/install.sh | bash, then supply credentials