Best MCP Servers for DevOps: GitHub, Docker, Cloudflare

DevOps teams spend a significant amount of time context-switching between GitHub, a container registry, a cloud console, and a monitoring dashboard. MCP servers let an AI agent hold that context for you, answering questions about deployment state, pipeline failures, and infrastructure configuration without requiring you to open four tabs. We analyzed the devtools category (3,141 servers) and the cloud category (164 servers) across MCPFind's index of 7,615 total MCP servers to identify the strongest options for CI/CD workflows, container management, and cloud infrastructure operations. For background on how MCP servers work, see what is MCP.

What Are the Best MCP Servers for DevOps and CI/CD in 2026?

MCPFind's devtools category is the largest in the directory at 3,141 servers, averaging 34.43 GitHub stars each. The cloud category adds 164 more, with the highest average star count of any category at 59.6 stars per server. Cloudflare leads the cloud category at 3,566 stars, followed by AWS, GCP, and platform-specific offerings from providers like Fly.io and Railway. On the devtools side, the top stars belong to developer tooling projects that extend well beyond CI/CD, but the category includes focused tools for GitHub, Docker, Kubernetes, Terraform, and CI platforms. We found the strongest DevOps MCP servers tend to come from the vendors themselves rather than community forks, because official servers have the most complete API coverage and the most consistent maintenance cadence. Browse all devtools servers to filter by use case.

How Does GitHub MCP Server Streamline Pull Request and Issue Management?

GitHub's official MCP server exposes pull requests, issues, repositories, code search, and Actions workflow status to your MCP client. In practical terms, that means an agent in Claude or Cursor can answer questions like "which PRs in the backend repo have been open for more than seven days without review" or "are there any failing Actions runs on main right now." Authentication uses a GitHub personal access token or a fine-grained token scoped to specific repositories. The fine-grained token option is the better production choice: it limits the agent's access to exactly the repositories and operations the workflow requires. The server runs in stdio mode and is open source, so you can audit the code before granting it access to your organization's repositories. For related DevOps workflows, our Docker deployment guide covers containerizing MCP servers for production use.

Which Cloud Infrastructure MCP Servers Have the Best Production Track Record?

Cloudflare MCP leads the cloud category at 3,566 stars, making it the highest-starred server in that category by a significant margin. It supports Workers deployment, D1 database queries, R2 storage management, KV namespace operations, and DNS record updates. Authentication uses a Cloudflare API token scoped to the specific zones or accounts you want the agent to access. Beyond Cloudflare, AWS has a growing collection of service-specific MCP servers covering EC2, S3, Lambda, and CloudWatch. GCP and Azure have similar collections, though with fewer community stars than Cloudflare's server. For teams that host MCP servers remotely, the remote hosting guide compares Cloudflare Workers, Cloud Run, and Vercel as deployment targets. All three options can host an MCP server with public HTTP endpoints while keeping secret management out of the client config.

How Do Docker and Kubernetes MCP Servers Fit Into Agentic Workflows?

Docker MCP comes directly from Docker Inc. and exposes build, run, push, and pull operations through stdio transport. An agent with Docker MCP configured can check running containers, inspect image layers, build from a Dockerfile, and push to a registry without leaving the AI interface. Authentication uses your existing Docker daemon credentials, which the server inherits through the local runtime. Kubernetes MCP handles cluster-level operations: listing pods, checking deployment status, reading logs, and applying manifests. Several community-maintained Kubernetes MCP servers exist in the devtools category, with authentication handled through kubeconfig file paths rather than stored tokens. Both tools work best in an interactive workflow where a developer reviews what the agent proposes before it executes destructive operations like container termination or rollout restarts. Read access can be used freely; write operations warrant a manual confirmation step.

Browse all DevOps options in the MCPFind devtools category or explore the full index.