Agentic CI/CD MCP Orchestrator MCP Server

Local setup required. This server has to be cloned and prepared on your machine before you register it in Claude Code.
1

Set the server up locally

Run this once to clone and prepare the server before adding it to Claude Code.

Run in terminal
pip install -r requirements.txt
2

Register it in Claude Code

After the local setup is done, run this command to point Claude Code at the built server.

Run in terminal
claude mcp add -e "OPENAI_API_KEY=${OPENAI_API_KEY}" -e "GITHUB_TOKEN=${GITHUB_TOKEN}" agentic-cicd-orchestrator -- node "<FULL_PATH_TO_MCPSERVER>/dist/index.js"

Replace <FULL_PATH_TO_MCPSERVER>/dist/index.js with the actual folder you prepared in step 1.

Required:OPENAI_API_KEYGITHUB_TOKEN+ 3 optional
README.md

GitHub Actions failure diagnosis and LLM-driven auto-repair PR creation.

Agentic CI/CD MCP Orchestrator

Python MCP server for GitHub Actions failure diagnosis, LLM-driven unified-diff auto-repair PR creation, and governed release orchestration.

What this provides

  • MCP tooling to inspect failed workflow runs using commit, logs, and test signals.
  • LLM diagnosis flow powered by OpenAI gpt-4o-mini.
  • Governance layer to auto-fix low-risk issues and require human review for risky changes.
  • Generic model-driven repair loop (up to 3 attempts) using unified diff patches.
  • GitHub Actions workflows for CI, repair orchestration, and release policy gating.

Project layout

  • mcp_server/main.py - MCP tools and orchestration entrypoints
  • mcp_server/run_repair.py - workflow-safe command runner
  • mcp_server/config.py - typed environment config
  • mcp_server/tools/* - GitHub, diagnosis, risk, and PR automation modules
  • .github/workflows/* - CI/CD automation workflows

Setup

  1. Create and activate a virtual environment.
  2. Install dependencies:
    • pip install -r requirements.txt
  3. Copy environment defaults:
    • cp .env.example .env (or create .env manually on Windows)
  4. Fill in required values (OPENAI_API_KEY, GITHUB_TOKEN).

Run MCP server locally

  • python -m mcp_server.main

Use in Cursor as MCP

  • MCP config is included at .cursor/mcp.json.
  • Restart Cursor so it loads the MCP server definition.
  • Ensure your .env has OPENAI_API_KEY and GITHUB_TOKEN.
  • In Cursor chat, call tools from agentic-cicd-orchestrator with:
    • repository: owner/repo
    • run_id: workflow run id (integer)
  • Main tools:
    • inspect_pipeline_failure
    • orchestrate_autofix

Run repair orchestration manually

  • Set REPOSITORY (for example org/repo).
  • Optional:
    • RUN_ID (if omitted, latest failed run is auto-selected)
    • WORKFLOW_NAME (filter latest failed run by workflow name, e.g. ci)
    • BASE_BRANCH (default main)
  • Execute: python -m mcp_server.run_repair

LLM auto-repair controls

  • MAX_REPAIR_ATTEMPTS - number of patch generation/application retries (default 3).
  • PATCH_STRATEGY - patch format expected from model (must be unified_diff).
  • LLM_PATCH_MAX_CHARS - upper bound on patch payload size.

Governance model

  • risk_score < RISK_AUTO_FIX_THRESHOLD -> autonomous auto-fix PR path.
  • RISK_AUTO_FIX_THRESHOLD <= risk_score < RISK_HUMAN_REVIEW_THRESHOLD -> human approval required.
  • risk_score >= RISK_HUMAN_REVIEW_THRESHOLD or high-risk file categories -> blocked/review-only path.
  • FORCE_AUTOFIX_ALL=true -> bypass thresholds and force auto-fix path (dangerous; use only in controlled testing).

Security notes

  • Use least-privilege GitHub credentials.
  • Keep production deployment credentials separate from auto-repair identity.
  • Review generated PRs and audit artifacts before enabling automerge in production.

Tools (2)

inspect_pipeline_failureInspects a failed GitHub Actions workflow run using commit, logs, and test signals.
orchestrate_autofixOrchestrates an autonomous fix or human-reviewed repair for a pipeline failure.

Environment Variables

OPENAI_API_KEYrequiredAPI key for OpenAI GPT-4o-mini model
GITHUB_TOKENrequiredGitHub personal access token for repository access
MAX_REPAIR_ATTEMPTSNumber of patch generation/application retries
PATCH_STRATEGYPatch format expected from model
RISK_AUTO_FIX_THRESHOLDThreshold for autonomous auto-fix PRs

Configuration

claude_desktop_config.json
{"mcpServers": {"agentic-cicd-orchestrator": {"command": "python", "args": ["-m", "mcp_server.main"], "env": {"OPENAI_API_KEY": "your-key", "GITHUB_TOKEN": "your-token"}}}}

Try it

→Inspect the latest failed workflow run for repository owner/repo and diagnose the issue.
→Orchestrate an autofix for the failed run ID 12345 in my repository.
→Check the logs for the recent CI failure in my project and suggest a repair patch.
→Analyze the risk score for the current pipeline failure and determine if it requires human review.

Frequently Asked Questions

What are the key features of Agentic CI/CD MCP Orchestrator?

Automated diagnosis of GitHub Actions workflow failures using LLMs. Generation of repair pull requests via unified diff patches. Governance layer to gate auto-fixes based on risk assessment thresholds. Support for multi-attempt repair loops. Integration with GitHub Actions for CI and release policy gating.

What can I use Agentic CI/CD MCP Orchestrator for?

Automatically patching minor syntax errors in CI test suites. Reducing developer toil by triaging and fixing common pipeline failures. Enforcing security governance on automated code changes in CI/CD pipelines. Scaling CI/CD maintenance by automating routine workflow repairs.

How do I install Agentic CI/CD MCP Orchestrator?

Install Agentic CI/CD MCP Orchestrator by running: pip install -r requirements.txt

What MCP clients work with Agentic CI/CD MCP Orchestrator?

Agentic CI/CD MCP Orchestrator works with any MCP-compatible client including Claude Desktop, Claude Code, Cursor, and other editors with MCP support.

Turn this server into reusable context

Keep Agentic CI/CD MCP Orchestrator docs, env vars, and workflow notes in Conare so your agent carries them across sessions.

Need the old visual installer? Open Conare IDE.
Open Conare
View all version-control servers →

Related MCP Servers

PR Review MCP Server
AI-assisted code review tool for developers/AQA engineers
version-control22 stars
Bitbucket Data Center
MCP server for Bitbucket Data Center to search code and manage pull requests.
version-control11 stars
Continuum
Git remembers what you changed. Continuum remembers why.
version-control6 stars
RepoCrunch
Analyze any public GitHub repository into structured JSON.
version-control5 stars
CERN GitLab MCP Server
Connects LLMs to CERN GitLab for discovering HEP code and documentation.
version-control3 stars
bbkt
A Bitbucket CLI and MCP server for managing workspaces and repositories.
version-control3 stars