Connects AI assistants to GitHub Actions, enabling developers and DevOps engineers to debug CI/CD pipelines with natural language queries. It diagnoses failures in workflow runs, analyzes logs to pinpoint errors and warnings, classifies root causes like dependency issues or syntax errors, and delivers targeted fix suggestions including code changes and YAML tweaks. Skip manual log dives and resolve issues faster.
Overview
An MCP server that connects AI assistants to GitHub Actions for CI/CD pipeline debugging. Diagnose failures, analyze logs, classify root causes, and get actionable fix suggestions — all through natural language.
Why Pipeline Doctor?
Debugging CI/CD pipelines is one of the most frustrating time sinks in software development. You navigate noisy UIs, parse verbose logs, and context-switch between dashboards and your IDE — often for the same categories of failures you've seen before.
Pipeline Doctor eliminates this workflow. It gives your AI assistant direct access to your pipeline data, a database of 88 failure patterns, and the ability to suggest fixes in seconds instead of minutes.
Features
- 9 tools for pipeline inspection, failure analysis, and workflow validation
- 5 prompts for guided debugging workflows (diagnose, health check, compare, validate, onboard)
- 3 resources for browsable workflow data and failure pattern catalog
- 88 failure patterns across 9 categories with context-aware fix suggestions
- Workflow validation with circular dependency detection, matrix resolution, and dry-run execution plans
- Mock mode for development and demos without a GitHub token
- Auto-detection of repository context from git remote
Tools
| Tool | Description |
|---|---|
| list_workflows | List all workflows in a repository with their current status and last run info |
| get_failed_runs | Fetch recent failed runs with branch, commit, trigger, duration, and actor |
| get_run_details | Get full structural breakdown of a run: jobs, steps, statuses, and timing |
| get_run_logs | Pull logs for a run with error extraction and smart truncation (50K char limit) |
| analyze_failure | Pattern-match logs against 88 failure patterns and return root cause analysis |
| suggest_fix | Generate actionable fix suggestions with corrected YAML snippets and commands |
| compare_runs | Diff two runs side-by-side: timing deltas, new vs resolved failures |
| validate_workflow | Validate workflow YAML: syntax, required fields, circular deps, deprecated commands |
| dry_run_workflow | Resolve matrix strategies and display the full execution plan without running jobs |
All tools accept an optional repo parameter in owner/repo format. When omitted, the repository is auto-detected from git remote or the --repo flag.
Prompts
Prompts are guided workflows that orchestrate multiple tools into a single interaction.
| Prompt | Description | Example Usage |
|---|---|---|
| diagnose | Full diagnosis of a failed run — logs, root cause, and fixes | "Diagnose run 12345 in my-org/my-repo" |
| pipeline-health | Health overview — workflow status and recent failures | "How healthy is my-org/my-repo's CI?" |
| compare | Compare two runs — timing changes and regressions | "Compare runs 12345 and 12346" |
| validate | Validate a workflow file and show its execution plan | "Validate my CI workflow" |
| onboard | Overview of a repo's CI/CD setup for new team members | "Show me the CI setup for my-org/my-repo" |
Resources
Resources provide browsable, read-only data that MCP clients can discover and subscribe to.
| Resource | URI | Description |
|---|---|---|
| Workflows | repo://{owner}/{repo}/workflows | Browsable listing of all workflows with current status |
| Failures | repo://{owner}/{repo}/failures | Recent failed runs (last 5) with metadata |
| Patterns | pipeline-doctor://patterns | Catalog of all 88 failure patterns grouped by category |
Use Cases
-
Build Failure Diagnosis: A developer asks, "Why did my latest Node.js build fail?" pipeline-doctor analyzes the logs from the failed GitHub Actions run, identifies a missing dependency, and suggests adding it to package.json.
-
Test Suite Debugging: For flaky tests in a Python workflow, query "Classify root cause of test failures in run #123." It categorizes as timeout issues and recommends increasing timeouts in the workflow YAML.
-
Deployment Rollback Analysis: After a failed deployment, input "Diagnose deployment failure and suggest fixes." It reviews logs, flags permission errors, and provides IAM policy adjustments.
-
Multi-Step Pipeline Troubleshooting: Query an entire workflow: "Analyze full pipeline logs for repo X and fix suggestions." Receives a comprehensive report with prioritized actions.
Who This Is For
Software developers building and maintaining GitHub-based CI/CD pipelines, DevOps engineers debugging automated workflows, and teams reducing manual log dives in fast-paced release cycles. Ideal for those integrating AI into development toolchains.
Links
Documentation: https://pipeline-doctor.pages.dev/
Contact: Rajib Dey