Top 8 MCP Servers for Cloud & DevOps in 2026
Cloud and DevOps is where Model Context Protocol (MCP) servers earn their keep — and where they carry the most risk. Connect an agent to one of these servers and it can list your Kubernetes pods, trigger a Netlify deploy, run any AWS CLI command, or sync a GitOps application, all from a natural-language prompt. That is enormously useful for on-call debugging, infrastructure inspection and release automation. It also means the server sits between an LLM and production credentials, which is exactly the kind of surface you want to vet before you wire it in.
Every server on this list has been scanned by MCP Trust Checker, so alongside the usual "is it popular and maintained?" you get an actual A–F Trust Score from a deterministic security analysis of the published package. Read the grade like a report card: A means the scan found the code clean and its capabilities well-scoped; B–F means there are things worth reviewing — over-broad permissions, or specific high-severity findings — before you grant it access to your cloud account. A high score is not a promise the server is harmless; it is a signal that the code did what it claimed and nothing alarming stood out. The number of high-severity findings is called out below wherever it is above zero.
The list is ordered best-first by a mix of real-world usefulness and grade. Popularity and capability lead — this is a "top servers" roundup, not a leaderboard sorted purely by score — but we flag every grade honestly so you can weigh convenience against risk yourself.
1. AWS API MCP Server (AWS Labs)
This is the official AWS Labs server, and the most direct way to drive AWS from an agent. It lets an LLM execute AWS CLI commands against your account — EC2, S3, IAM, Lambda, CloudFormation and effectively the entire surface — with command validation, IAM-scoped credentials, and an optional read-only mode for safe inspection. It ships as part of the widely-starred awslabs/mcp monorepo of purpose-built AWS servers. Because it can reach so much of your account, the read-only mode and tight IAM scoping matter: use them.
Our scan flags 1 high-severity finding — review it on the registry page before granting write access to a production account.
MCP Trust Score: A (98/100) · GitHub · PyPI
2. Kubernetes MCP Server (Flux159)
One of the most popular Kubernetes MCP servers on npm and a near-default in DevOps roundups. It connects to any cluster through your local kubectl context and lets agents list and describe pods, deployments and services, read logs, apply manifests, and run Helm operations. It leans toward safe read operations by default to avoid accidentally mutating cluster state, and works with EKS, GKE, AKS or a local cluster. Ideal for turning an agent into an on-call assistant that can actually explain why a pod is crash-looping.
Be aware: our scan flags 5 high-severity findings — the most on this list. The grade is still an A on overall structure, but read those findings on its registry page and prefer running it against a read-scoped context before you let an agent apply manifests to production.
MCP Trust Score: A (94/100) · GitHub · npm
3. Docker MCP Server
A lightweight Python server for local Docker operations. It lets an agent create and run containers, deploy Docker Compose stacks, pull logs, and inspect running container state — handy for spinning up local environments or automating quick container ops inside an AI workflow. It fills the container-management slot in a DevOps stack without the weight of a full orchestration integration.
Our scan flags 1 high-severity finding; it also carries a high capability level because container control is powerful by nature. Review the finding on its registry page before wiring it into anything shared.
MCP Trust Score: A (98/100) · GitHub · PyPI
4. Argo CD MCP Server
Maintained under argoproj-labs, the Argo project org, this is the go-to server for GitOps and continuous delivery. Agents can list clusters and applications, create, update and delete apps, sync them, and manage application resources conversationally. It supports both stdio and HTTP transports, so it fits local and hosted setups alike, and it brings the CD loop into reach of an agent that is already reasoning about your deployments.
This one earns a perfect score with zero high-severity findings — a clean scan and a minimal capability footprint.
MCP Trust Score: A (100/100) · GitHub · npm
5. CircleCI MCP Server
CircleCI's official server, published and maintained by CircleCI-Public. It connects an agent directly to your CircleCI project state so it can inspect pipelines and jobs, pull build and test failure logs, and reason about CI results in plain language. The payoff is fast: instead of clicking through a failed pipeline, you ask why the build broke and get the relevant logs summarized. A strong pick for anyone whose team already lives in CircleCI.
Another perfect score with no high-severity findings and a minimal capability level — largely because it reads CI state rather than mutating infrastructure.
MCP Trust Score: A (100/100) · GitHub · npm
6. Prometheus MCP Server
The most recommended standalone Prometheus server, and a favourite in SRE and DevOps lists. It gives an LLM direct, structured access to your metrics backend: run PromQL queries, discover metrics and metadata, and analyze time-series for troubleshooting and capacity work — no dashboard round-trip required. It plugs into Claude Desktop, Cursor, VS Code and other clients, making it a natural companion to the Kubernetes and cloud servers above when you are chasing down an incident.
Zero high-severity findings and a minimal capability footprint — it queries metrics, it doesn't touch infrastructure, which is reflected in the clean scan.
MCP Trust Score: A (98/100) · GitHub · PyPI
7. Netlify MCP Server
Netlify's official server wraps its API and CLI so agents can create projects, trigger and monitor deploys, manage environment variables, and inspect site and build resources. It is a clean fit for web-app deployment automation and edge/serverless functions, and it is actively maintained with frequent releases. If your delivery target is Netlify, this turns "ship this and tell me if the build fails" into a single agent instruction.
Our scan flags 1 high-severity finding and a high capability level (it can manage environment variables and deploys). Review the finding on its registry page before connecting it to a production site.
MCP Trust Score: A (98/100) · GitHub · npm
8. Azure MCP Server
Microsoft's official MCP server for Azure, and the standard integration for the second-largest cloud. It exposes resource groups, storage, Cosmos DB, the Azure CLI, and Monitor / Log Analytics (KQL) to agents, ships an azmcp binary, and integrates cleanly with GitHub Copilot for Azure in VS Code. If your infrastructure lives in Azure, it is the obvious choice for querying and managing resources conversationally.
Honesty first: this is the only server here below an A. Our scan gives it a B (87/100) with 1 high-severity finding and a high capability level — expected for an official, broad-surface cloud server, but worth reviewing on its registry page before you grant it write access. The B is a "read the findings and scope your credentials" flag, not a reason to avoid it.
MCP Trust Score: B (87/100) · GitHub · npm
How we graded these
Every Trust Score above comes from the deterministic, open-source MCP Trust Checker engine, which reads the actual published npm or PyPI source of each server and applies a fixed set of security rules — the Capability-Flow Trust Model and toxic-flow analysis. There is no LLM in the loop and no vibes: the same input package always produces the same score, so the grades are reproducible and auditable rather than subjective. Higher-capability servers (those that can mutate cloud state) are held to a stricter bar, which is why a broad official server can score lower than a focused read-only one. You can see the full findings behind any grade on the registry, and read how the scoring works on the security methodology page.
Want the complete picture? Browse every graded server in this category at the Cloud & DevOps registry, and scan any package you're considering — including one that isn't listed here — for free via the MCP Trust Checker API. Check the score before you connect the server, not after.
Frequently asked questions
What is the best MCP server for cloud and DevOps in 2026?
For AWS, the official AWS Labs AWS API MCP Server is the definitive choice (Trust Score A, 98/100). For Kubernetes, the Flux159 mcp-server-kubernetes is the most popular. The 'best' depends on your stack — but among the highest-graded on our registry, Argo CD (GitOps) and CircleCI (CI) both score a perfect A (100/100) with zero high-severity findings. Compare them all at https://mcptrustchecker.com/registry/category/cloud-devops.
Are cloud and DevOps MCP servers safe to use?
They can be, but they're high-risk by nature because they sit between an AI agent and production credentials. That's why every server on our list carries an A–F MCP Trust Score from a deterministic security scan. Prefer servers with high grades and zero high-severity findings, use read-only modes and IAM-scoped credentials, and review the findings on each server's registry page before granting write access.
What does the MCP Trust Score mean?
It's a reproducible A–F (0–100) grade from the open-source MCP Trust Checker engine, which reads a server's published npm or PyPI source and applies fixed security rules — no LLM involved, so the same package always yields the same score. A means a clean scan with well-scoped capabilities; B–F means there are findings worth reviewing, such as over-broad permissions or high-severity issues.
Which cloud MCP servers have the highest Trust Score?
On our registry, the Argo CD MCP Server and the CircleCI MCP Server both score A (100/100) with zero high-severity findings. AWS API, Docker, Prometheus and Netlify each score A (98/100). The Azure MCP Server scores B (87/100) with one high-severity finding — still usable, but review it first.
How can I check an MCP server's security before installing it?
Look it up on the MCP Trust Registry at https://mcptrustchecker.com/registry, or scan any npm/PyPI package for free through the API at https://mcptrustchecker.com/api. The scan is deterministic and reads the real published source, so you get the same reproducible grade every time — check it before you connect the server to your cloud account.
Scan your MCP server now
MCP Trust Checker is free, open-source and runs entirely on your machine. Get an A–F Trust Score for any MCP server in seconds.
npx mcptrustchecker