Official
Query dashboards, Prometheus/Loki datasources, alerts, and incidents from an AI agent — official.
grafana, observability, prometheus, loki
Best Cloud & Infrastructure MCP Servers in 2026
Cloud and infrastructure MCPs for AWS, GCP, Azure, and edge platforms. Let agents manage deployments and infra — verified for 2026.
Top Cloud & Infrastructure MCPs
- 1.Grafana—Query dashboards, Prometheus/Loki datasources, alerts, and incidents from an AI agent — official.
- 2.Terraform—Look up Terraform Registry providers and modules, and manage HCP Terraform workspaces from an AI agent.
- 3.Azure—Manage Azure resources across 40+ services from an AI agent — Microsoft's official MCP.
Ranked by top-mcps.com — data verified Sep 2026.
About Cloud & Infrastructure MCP servers
The best MCP servers for cloud and infrastructure expose the consoles a platform team lives in — AWS, Azure, Cloudflare, DigitalOcean, Kubernetes, Vercel — to AI agents that can read state, suggest fixes, and, with explicit permission, make changes. The good ones draw a hard line between read and write capabilities and surface a clear 'this would do X' preview before any mutation, because the distance between 'inspect the failing deploy' and 'delete the resource' is one tool call.
Choose by surface. For deploy status, build logs, and env-var audits, Vercel (MCP server) is an api-key stdio install of about five minutes. DigitalOcean (api-key) covers App Platform deploys, Droplets, and DOKS clusters; Supabase (OAuth) manages database, auth, storage, and edge functions for teams whose whole backend lives there; Stripe sits at the billing edge — inspecting subscription state and issuing refunds with approval. Among the big clouds, Azure is Microsoft's official server with OAuth 2.1 across 40+ services including Monitor and Log Analytics; AWS uses api-key credentials and earns its fifteen-minute setup with CloudWatch log search and IAM audits; Google Cloud Storage (OAuth 2.1, official) reads and writes bucket objects. Cloudflare (OAuth 2.1, stdio or remote streamable-http) handles Workers, DNS, and R2. For clusters, the Kubernetes MCP needs no credential of its own — it inherits your kubeconfig, which is exactly why you point it at a dev cluster and use its non-destructive mode. For infrastructure-as-code, Terraform grounds generated HCL in real provider schemas and module inputs — registry lookups need no auth at all. Grafana rounds out incident work by querying Prometheus and Loki through one server.
Common mistakes: granting the MCP an admin role when read-only would do, running it with a long-lived static credential instead of a scoped short-lived one, and ignoring API rate limits — an agent that retries aggressively will trip AWS throttling mid-incident, exactly when you need the calls to land. The classic disaster is context confusion: a kubectl-equivalent call that ran against production because that was the kubeconfig default. Keep plan and apply separated in Terraform workflows — the agent drafts, a human applies. Start with AWS (or your primary cloud's MCP) in read-only mode against a non-production account, prove one investigation end to end, then add write capabilities under change management.
All Cloud & Infrastructure MCPs
11 MCPs ranked by popularity. Filter by attribute or search by name.
11 of 11 MCPs
| # | MCP | Tags | Setup | Complexity | Labels | |
|---|---|---|---|---|---|---|
| 1 | Grafana Query dashboards, Prometheus/Loki datasources, alerts, and incidents from an AI agent — official. | grafana, observability | 5 min | Medium | Official | |
| 2 | Terraform Look up Terraform Registry providers and modules, and manage HCP Terraform workspaces from an AI agent. | terraform, iac | 5 min | Medium | Official | |
| 3 | Azure Manage Azure resources across 40+ services from an AI agent — Microsoft's official MCP. | azure, microsoft | 5 min | Medium | Official | |
| 4 | Stripe Manage payments, customers, and subscriptions through Stripe. | stripe, payments | 5 min | Medium | Official | |
| 5 | Cloudflare Manage Workers, R2 buckets, DNS, and edge policies on Cloudflare. | cloudflare, edge | 10 min | Medium | Official | |
| 6 | Google Cloud Storage Read, write, and manage GCS buckets and objects from an AI agent — official Google MCP, local or remote. | google-cloud, gcs | 10 min | Medium | Official | |
| 7 | Supabase Full Supabase access: database, auth, storage, and edge functions. | supabase, database | 5 min | Low | ||
| 8 | AWS AWS's own managed MCP server — run AWS APIs, execute sandboxed scripts, and search AWS documentation under IAM. | aws, cloud | 10 min | Medium | Official | |
| 9 | Kubernetes Inspect and manage Kubernetes clusters through kubectl-equivalent tools, with a non-destructive mode. | kubernetes, k8s | 5 min | Medium | ||
| 10 | DigitalOcean Manage Droplets, App Platform, Kubernetes, and Spaces from an AI agent. | digitalocean, hosting | 5 min | Medium | Official | |
| 11 | Vercel (MCP server) Inspect deployments, logs, and project state on Vercel from an AI agent. | vercel, hosting | 5 min | Low | Official |
Choose the right MCP
Quick decision guide based on your use case.
| If you need… | Start with |
|---|---|
| You deploy on Vercel and debug builds often | Use Vercel (MCP server) |
| You are on AWS and investigate incidents | Use AWS with a read-only IAM principal |
| Your org runs on Microsoft Azure | Use Azure |
| You manage DNS, Workers, or R2 at the edge | Use Cloudflare |
| An agent drafts your infrastructure-as-code | Use Terraform; keep apply behind human review |
| Pods are failing and you need fast diagnosis | Use Kubernetes in non-destructive mode |
| Your whole backend runs on Supabase | Use Supabase |
Top Cloud & Infrastructure MCPs ranked
Detailed cards with setup time, complexity, and key labels.
Official
Look up Terraform Registry providers and modules, and manage HCP Terraform workspaces from an AI agent.
terraform, iac, infrastructure-as-code, hashicorp
Official
Manage Azure resources across 40+ services from an AI agent — Microsoft's official MCP.
azure, microsoft, cloud, aks
Official
Manage payments, customers, and subscriptions through Stripe.
stripe, payments, billing, invoicing
Official
Manage Workers, R2 buckets, DNS, and edge policies on Cloudflare.
cloudflare, edge, workers, dns
Official
Read, write, and manage GCS buckets and objects from an AI agent — official Google MCP, local or remote.
google-cloud, gcs, storage, object-storage
Full Supabase access: database, auth, storage, and edge functions.
supabase, database, backend, postgres
8
OfficialAWS's own managed MCP server — run AWS APIs, execute sandboxed scripts, and search AWS documentation under IAM.
aws, cloud, cloudwatch, iam
Inspect and manage Kubernetes clusters through kubectl-equivalent tools, with a non-destructive mode.
kubernetes, k8s, kubectl, helm
Official
Manage Droplets, App Platform, Kubernetes, and Spaces from an AI agent.
digitalocean, hosting, cloud, kubernetes
Official
Inspect deployments, logs, and project state on Vercel from an AI agent.
vercel, hosting, deployment, nextjs
Archived (historical reference)
1 Cloud & Infrastructure entry is archived — the upstream package was deprecated or pulled, or a documented security issue applies. The detail page is preserved for historical reference and migration guidance, but these are NOT current editorial picks.
Also in the ChatGPT Apps directory
1 Cloud & Infrastructure brand is available as one-click ChatGPT Apps. These are not ranked alongside the editorial picks above — they're listings from chatgpt.com/apps surfaced here for brand-search continuity.
FAQ: Cloud & Infrastructure MCPs
Which cloud MCP do I install first?
AWS for most teams (it is the dominant cloud), Cloudflare for sites and edge workers, Vercel for Next.js deploys. Pair the cloud MCP with a dedicated, scoped IAM principal — never your personal admin role.
Should I give an agent terraform plan + apply access?
No. Plan-only is the safe default. Apply requires human review of the diff — agents drafting infra is high-leverage, agents executing infra unsupervised is high-risk.
Can I let an agent debug my Kubernetes cluster safely?
Yes, with two controls: run the Kubernetes MCP in its non-destructive mode so it can read pods, logs, and events but not delete anything, and remember it inherits your kubeconfig — the classic failure is a command that ran against production because that context was the default. Pin the context to a dev cluster or a single namespace before wiring it up.
Which MCP helps write Terraform rather than run it?
The Terraform MCP. It grounds generated HCL in real provider schemas and looks up module inputs and outputs from the registry, which kills the most common failure in agent-written IaC: hallucinated resource arguments. Registry lookups need no auth at all; connecting HCP Terraform for workspace and run management is optional. Keep apply behind human review either way.
How do cloud MCPs authenticate, and which model is safer?
Azure, Cloudflare, Google Cloud Storage, and Supabase use OAuth 2.1; AWS, Vercel (MCP server), DigitalOcean, Grafana, and Stripe use api-keys or tokens. OAuth flows take more setup but produce scoped, revocable grants; api-keys are faster but become standing credentials. Either way, create a dedicated principal per MCP with the smallest role that works — never your personal admin — and rotate keys on a schedule.