GitLab-CICD-Agent
agentic-cicd is an AI-powered multi-agent system that automates GitLab CI/CD workflows using natural language. Built with LangGraphJS, it connects to GitLab via OAuth, interprets user intent, generates pipelines, and executes deployments — all orchestrated by autonomous AI agents and backed by a GitLab MCP server.
claude mcp add --transport stdio shalwin04-gitlab-cicd-agent node path/to/server.js \ --env SUPABASE_KEY="your_supabase_key" \ --env SUPABASE_URL="your_supabase_url" \ --env TAVUS_API_KEY="your_tavus_key" \ --env GOOGLE_API_KEY="your_gemini_api_key" \ --env GITLAB_CLIENT_ID="your_gitlab_client_id" \ --env GITLAB_REDIRECT_URI="https://your-frontend-url.vercel.app/oauth/callback" \ --env GITLAB_CLIENT_SECRET="your_gitlab_client_secret"
How to use
The GitLab-CICD-Agent MCP server provides a natural-language driven interface to automate GitLab CI/CD workflows. It leverages a multi-agent system to interpret user requests, generate the necessary CI/CD artifacts (such as .gitlab-ci.yml, Dockerfiles, and deployment manifests), and push changes to GitLab while opening merge requests. You interact with the system by describing your desired pipeline in plain language, and the Supervisor and sub-agents handle plan generation, code/artifact creation, validation, and integration with GitLab via the dedicated MCP Tool Server.
To get started, connect your GitLab account through the app, then issue commands like: “Set up CI to run tests on push and deploy to GKE when main is updated.” The planner agent will break down the task, the code agent will generate the necessary GitLab CI configuration and infrastructure files, and the test agent will validate them. The GitLab MCP Server will apply changes (open MR, trigger pipelines), and you can monitor progress from the UI or GitLab itself. You can ask the ChatAgent for explanations of what the pipeline is doing or why a build failed.
How to install
Prerequisites: Node.js and npm installed on your machine; access to a GitLab account; optional: Git for cloning and pushing to repositories.
- Clone the repository
git clone https://github.com/your-org/agentic-cicd.git
cd agentic-cicd
- Install dependencies for backend and (if applicable) frontend
# Install backend dependencies
cd backend
npm install
# If you also run the frontend locally
cd ../frontend
npm install
- Create environment configuration
- Backend/.env (example placeholders)
GITLAB_CLIENT_ID=your_gitlab_client_id
GITLAB_CLIENT_SECRET=your_gitlab_client_secret
GITLAB_REDIRECT_URI=https://your-frontend-url.vercel.app/oauth/callback
SUPABASE_URL=your_supabase_url
SUPABASE_KEY=your_supabase_key
google_api_key=your_gemini_api_key
TAVUS_API_KEY=your_tavus_key
- Run the MCP server (node-based)
# From repository root or where your server entrypoint resides
npm run build-helper-if-needed
# Start the MCP server (adjust path as needed)
node path/to/server.js
- Optional: run frontend/backend locally if needed
# Frontend
cd frontend
npm run dev
# Backend
cd backend
npm run start
Additional notes
Environment variables listed in the mcp_config block should be provided securely in your deployment environment. Ensure GitLab OAuth scopes grant access to the repositories you intend to manage (API read/write, MR creation, pipeline triggering). The system generates artifacts like .gitlab-ci.yml and Dockerfiles; review them in a feature branch before opening MRs. If you encounter YAML syntax errors during validation, use the TestAgent to perform linting and a dry-run of GitLab CI if supported. Monitor pipeline status directly in GitLab or via the ChatAgent updates in the MCP UI. For large deployments, consider enabling observability and tracing in your MCP Tool Server for easier debugging.
Related MCP Servers
gemini-cli
An open-source AI agent that brings the power of Gemini directly into your terminal.
agent
A unified Model Context Protocol server implementation that aggregates multiple MCP servers into one.
pluggedin-app
The Crossroads for AI Data Exchanges. A unified, self-hostable web interface for discovering, configuring, and managing Model Context Protocol (MCP) servers—bringing together AI tools, workspaces, prompts, and logs from multiple MCP sources (Claude, Cursor, etc.) under one roof.
furi
CLI & API for MCP management
create -app
A CLI tool for quickly scaffolding Model Context Protocol (MCP) server applications with TypeScript support and modern development tooling
biznagafest
MCP Servers en Málaga con salero