Cloud Provider MCP Servers: AWS, Azure & Google Cloud

Cloud provider MCP servers connect AI applications to the APIs of major cloud platforms -- AWS, Azure, Google Cloud, and Cloudflare. They enable AI assistants to manage infrastructure, query cloud services, optimize costs, and automate DevOps workflows through natural language conversations.

This guide covers the MCP server landscape for every major cloud provider, including setup instructions, available tools, security best practices, and practical workflow examples for AI-powered cloud management.

Why Cloud Provider MCP Servers Matter

Managing cloud infrastructure is inherently complex. Cloud platforms offer hundreds of services, each with its own API, configuration options, and best practices. Cloud MCP servers address this by letting AI assistants:

• Query infrastructure state: "What EC2 instances are running in us-east-1?"
• Troubleshoot issues: "Why is my Lambda function timing out?"
• Optimize costs: "Which instances are underutilized and could be downsized?"
• Generate IaC code: "Create a Terraform configuration for this architecture"
• Monitor and alert: "Show me the CloudWatch metrics for my API Gateway"

Instead of navigating console UIs or writing API calls, you describe what you need in natural language.

AWS MCP Servers

AWS offers the most extensive MCP server ecosystem, with both official AWS-maintained servers and a rich community ecosystem.

Official AWS MCP Servers

AWS has released official MCP servers covering several core services:

AWS CDK MCP Server Setup

The AWS CDK MCP server is particularly powerful for infrastructure-as-code workflows:

{
  "mcpServers": {
    "aws-cdk": {
      "command": "npx",
      "args": ["-y", "@aws/mcp-server-cdk"],
      "env": {
        "AWS_PROFILE": "your-profile",
        "AWS_REGION": "us-east-1"
      }
    }
  }
}

Available Tools:

Example Workflow:

User: "Create a CDK stack for a serverless API with API Gateway,
       Lambda, and DynamoDB"

Claude's workflow:
1. SearchConstructs("serverless API") — find relevant patterns
2. GenerateCDKConstruct(description) — generate the construct code
3. ValidateCDKStack(code) — validate the generated stack
4. Output the complete CDK TypeScript code for review

Amazon S3 MCP Server

{
  "mcpServers": {
    "aws-s3": {
      "command": "npx",
      "args": ["-y", "@aws/mcp-server-s3"],
      "env": {
        "AWS_ACCESS_KEY_ID": "your_key",
        "AWS_SECRET_ACCESS_KEY": "your_secret",
        "AWS_REGION": "us-east-1"
      }
    }
  }
}

Capabilities:

• List buckets and objects
• Upload and download files
• Manage bucket policies and lifecycle rules
• Generate presigned URLs
• Query storage metrics and costs

AWS Lambda MCP Server

{
  "mcpServers": {
    "aws-lambda": {
      "command": "npx",
      "args": ["-y", "@aws/mcp-server-lambda"],
      "env": {
        "AWS_PROFILE": "your-profile",
        "AWS_REGION": "us-east-1"
      }
    }
  }
}

Capabilities:

• List and describe Lambda functions
• View function configurations and environment variables
• Read CloudWatch logs for function executions
• Invoke functions with test payloads
• Update function code and configurations

Community AWS MCP Servers

The community has built additional AWS MCP servers covering services not yet available officially:

Azure MCP Servers

Microsoft Azure's MCP server ecosystem provides access to Azure services for AI-powered management:

Azure MCP Server Setup

{
  "mcpServers": {
    "azure": {
      "command": "npx",
      "args": ["-y", "mcp-server-azure"],
      "env": {
        "AZURE_SUBSCRIPTION_ID": "your-subscription-id",
        "AZURE_TENANT_ID": "your-tenant-id",
        "AZURE_CLIENT_ID": "your-client-id",
        "AZURE_CLIENT_SECRET": "your-client-secret"
      }
    }
  }
}

Available Azure Tools

Azure DevOps MCP Server

For teams using Azure DevOps, a dedicated MCP server provides access to pipelines, boards, and repos:

{
  "mcpServers": {
    "azure-devops": {
      "command": "npx",
      "args": ["-y", "mcp-server-azure-devops"],
      "env": {
        "AZURE_DEVOPS_ORG": "your-org",
        "AZURE_DEVOPS_PAT": "your-personal-access-token"
      }
    }
  }
}

Capabilities:

• Manage work items and boards
• View and trigger build/release pipelines
• Access repository operations (similar to GitHub MCP)
• Query test results and code coverage

Google Cloud MCP Servers

Google Cloud MCP servers provide access to GCP services:

Google Cloud MCP Server Setup

{
  "mcpServers": {
    "gcp": {
      "command": "npx",
      "args": ["-y", "mcp-server-gcp"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/service-account.json",
        "GCP_PROJECT_ID": "your-project-id"
      }
    }
  }
}

GCP Service Coverage

BigQuery MCP Server

BigQuery deserves special attention as a standalone MCP server for data analytics:

{
  "mcpServers": {
    "bigquery": {
      "command": "npx",
      "args": ["-y", "mcp-server-bigquery"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/service-account.json",
        "GCP_PROJECT_ID": "your-project-id"
      }
    }
  }
}

Capabilities:

• Execute SQL queries against BigQuery datasets
• List datasets, tables, and schemas
• View query costs before execution
• Export query results
• Manage table schemas and partitioning

Example Workflow:

User: "Analyze our website traffic data in BigQuery for the last 30 days"

Claude's workflow:
1. list_datasets() — discover available datasets
2. list_tables("analytics") — find traffic-related tables
3. describe_table("analytics.events") — understand the schema
4. query("SELECT event_name, COUNT(*) as count,
         COUNT(DISTINCT user_id) as unique_users
         FROM analytics.events
         WHERE event_date >= DATE_SUB(CURRENT_DATE(), INTERVAL 30 DAY)
         GROUP BY event_name
         ORDER BY count DESC
         LIMIT 20") — execute the analysis

Cloudflare MCP Server

Cloudflare provides an official MCP server for managing its edge platform:

{
  "mcpServers": {
    "cloudflare": {
      "command": "npx",
      "args": ["-y", "@cloudflare/mcp-server-cloudflare"],
      "env": {
        "CLOUDFLARE_API_TOKEN": "your_api_token"
      }
    }
  }
}

Cloudflare Tools

Cloud Provider Comparison

Security Best Practices

IAM Configuration

Create dedicated, limited-scope credentials for MCP servers:

AWS Example (IAM Policy):

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "MCPReadOnly",
      "Effect": "Allow",
      "Action": [
        "ec2:Describe*",
        "s3:List*",
        "s3:GetObject",
        "lambda:List*",
        "lambda:GetFunction",
        "cloudwatch:GetMetricData",
        "cloudwatch:ListMetrics",
        "ce:GetCostAndUsage"
      ],
      "Resource": "*"
    },
    {
      "Sid": "DenyDangerousActions",
      "Effect": "Deny",
      "Action": [
        "ec2:TerminateInstances",
        "s3:DeleteBucket",
        "iam:*",
        "organizations:*"
      ],
      "Resource": "*"
    }
  ]
}

GCP Example (Custom Role):

title: MCP Read Only
description: Limited access for MCP server
stage: GA
includedPermissions:
  - compute.instances.list
  - compute.instances.get
  - storage.buckets.list
  - storage.objects.list
  - storage.objects.get
  - bigquery.datasets.get
  - bigquery.tables.list
  - bigquery.jobs.create
  - logging.logEntries.list

Network Security

• Run MCP servers within your VPC/VNet when possible
• Use VPC endpoints or Private Link for API access
• Restrict outbound access to only necessary cloud API endpoints
• Enable VPC flow logs for network monitoring

Audit and Compliance

• Enable CloudTrail (AWS), Activity Log (Azure), or Audit Logs (GCP) to track all API calls made by MCP servers
• Set up alerts for sensitive operations (resource creation, permission changes)
• Review logs regularly to ensure the AI is not making unexpected API calls
• For compliance requirements, see our MCP Security & Compliance guide

Common Workflows

Infrastructure Discovery and Documentation

User: "Document all the resources in our AWS production account"

Claude's workflow:
1. Queries each service (EC2, RDS, S3, Lambda, etc.)
2. Catalogs all resources with their configurations
3. Maps relationships (which Lambda uses which DynamoDB table)
4. Generates architecture documentation
5. Identifies resources without proper tags

Cost Optimization

User: "Find ways to reduce our AWS bill"

Claude's workflow:
1. Query Cost Explorer for spending by service
2. Identify top cost drivers
3. Check EC2 instances for utilization metrics
4. Find unused EBS volumes and unattached Elastic IPs
5. Analyze S3 storage classes and lifecycle rules
6. Recommend specific optimizations with estimated savings

Multi-Cloud Correlation

When running MCP servers for multiple providers:

User: "Compare our compute costs across AWS and GCP"

Claude's workflow:
1. AWS Cost Explorer: Get EC2 spending breakdown
2. GCP Billing: Get Compute Engine spending breakdown
3. Normalize instance types to comparable tiers
4. Present side-by-side comparison with recommendations

Incident Response

User: "Our API is returning 5xx errors. Help me debug."

Claude's workflow:
1. CloudWatch: Check API Gateway error metrics
2. Lambda: Pull recent function invocation logs
3. RDS: Check database connection count and CPU
4. Identify root cause (e.g., database connection exhaustion)
5. Suggest remediation steps

Deploying MCP Servers for Cloud Management

For organizations wanting centralized MCP access to cloud resources, consider deploying MCP servers as remote services:

Docker Deployment

FROM node:20-slim
RUN npm install -g @aws/mcp-server-s3 @aws/mcp-server-lambda
EXPOSE 3000
CMD ["node", "server.js"]

Kubernetes Deployment

apiVersion: apps/v1
kind: Deployment
metadata:
  name: cloud-mcp-servers
spec:
  replicas: 2
  template:
    spec:
      containers:
        - name: aws-mcp
          image: your-registry/aws-mcp-server
          envFrom:
            - secretRef:
                name: aws-credentials
          resources:
            limits:
              memory: 256Mi
              cpu: 250m

For detailed deployment guides, see Deploying Remote MCP Servers.

Integration with IaC Tools

Cloud MCP servers work alongside Infrastructure as Code tools:

The recommended pattern: AI proposes infrastructure changes as code, humans review and approve, then automated pipelines apply the changes.

Cloud Provider MCP Server Selection Guide

Decision Matrix

Getting Started Checklist

1. Choose your primary cloud provider MCP server
2. Create a dedicated IAM user/role with minimum permissions
3. Configure credentials using environment variables (never hardcode)
4. Test read-only operations first (list, describe, get)
5. Add monitoring/alerting MCP server if available
6. Gradually expand permissions as needed for specific workflows
7. Enable audit logging before granting any write access
8. Set up cost alerts to monitor API usage spending

Real-World Cloud Management Workflows

Workflow: Cloud Migration Assessment

AI-powered analysis of migration readiness:

User: "Assess our AWS infrastructure for migration to a
       containerized architecture"

Claude's workflow:
1. (AWS EC2) List all running instances with configurations
2. (AWS RDS) Catalog database instances and their dependencies
3. (AWS ELB) Map load balancer configurations
4. (AWS CloudWatch) Analyze resource utilization patterns
5. Assess each service for containerization readiness:
   - Stateless services → Direct container migration
   - Stateful services → Requires data migration planning
   - Legacy services → May need refactoring
6. Generate migration plan with phases, risks, and timeline

Workflow: Security Posture Review

User: "Audit our cloud security configuration"

Claude's workflow:
1. (AWS IAM) List users, roles, and policies
2. (AWS EC2) Check security group configurations
3. (AWS S3) Verify bucket policies and public access settings
4. (AWS CloudTrail) Verify logging is enabled
5. Flag issues:
   - IAM users without MFA enabled
   - Security groups with 0.0.0.0/0 ingress rules
   - S3 buckets with public read access
   - Unused IAM access keys older than 90 days
6. Generate a prioritized remediation checklist

Workflow: Disaster Recovery Testing

User: "Verify our disaster recovery setup for the payment service"

Claude's workflow:
1. (AWS RDS) Check automated backup configuration
2. (AWS RDS) Verify cross-region replication status
3. (AWS Route53) Check failover DNS configuration
4. (AWS S3) Verify backup bucket replication
5. (AWS Lambda) Check DR automation functions
6. Generate DR readiness report with any gaps identified

Cost Management Deep Dive

Cloud cost optimization is one of the most valuable applications of cloud MCP servers. Here are detailed strategies:

Rightsizing Analysis

User: "Analyze our EC2 instances for rightsizing opportunities"

Claude's workflow:
1. (AWS EC2) List all instances with types and tags
2. (AWS CloudWatch) Get CPU, memory, network metrics for each
3. Analyze utilization patterns:
   - Instances consistently under 20% CPU → candidates for downsize
   - Instances with burst patterns → switch to burstable types
   - Instances unused on weekends → add scheduling
4. Calculate potential savings per recommendation
5. Generate report with estimated monthly savings

Reserved Instance and Savings Plan Analysis

AI assistants can analyze your usage patterns and recommend the optimal mix of commitment types.

Waste Identification

Common sources of cloud waste that AI can identify:

Multi-Cloud Architecture Patterns

Unified Multi-Cloud Dashboard

Configure MCP servers for all your cloud providers in a single session:

{
  "mcpServers": {
    "aws": {
      "command": "npx",
      "args": ["-y", "@aws/mcp-server-ec2"],
      "env": { "AWS_PROFILE": "production" }
    },
    "gcp": {
      "command": "npx",
      "args": ["-y", "mcp-server-gcp"],
      "env": { "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/sa.json" }
    },
    "azure": {
      "command": "npx",
      "args": ["-y", "mcp-server-azure"],
      "env": { "AZURE_SUBSCRIPTION_ID": "xxx" }
    },
    "cloudflare": {
      "command": "npx",
      "args": ["-y", "@cloudflare/mcp-server-cloudflare"],
      "env": { "CLOUDFLARE_API_TOKEN": "xxx" }
    }
  }
}

This enables cross-provider queries and comparisons in a single conversation.

Cloud-Agnostic Patterns

When building cloud-agnostic infrastructure with AI assistance:

1. Service mapping: AI understands equivalent services across providers (AWS Lambda / Cloud Functions / Azure Functions)
2. Terraform generation: AI generates provider-agnostic Terraform modules
3. Cost comparison: AI compares pricing across providers for similar workloads
4. Migration planning: AI assists with migration between cloud providers

Troubleshooting Cloud MCP Servers

Common Issues

Debugging Checklist

1. Verify credentials are set correctly in environment variables
2. Test credentials manually with the CLI (aws sts get-caller-identity)
3. Check that the IAM user/role has the necessary permissions
4. Verify network connectivity to the cloud provider's API endpoints
5. Check for any service-specific prerequisites (e.g., bucket must exist for S3 operations)
6. Review MCP server logs for specific error messages

Serverless Function Management

Cloud MCP servers are particularly effective for managing serverless architectures, where the number of individual resources can be large and their interactions complex.

Lambda Function Lifecycle Management

AI assistants with access to AWS Lambda MCP servers can streamline the entire serverless lifecycle:

User: "List all Lambda functions that haven't been invoked in the
       last 30 days and estimate the cost savings from removing them"

Claude's workflow:
1. (AWS Lambda) list_functions() → all functions with metadata
2. (AWS CloudWatch) For each function, check invocation count
   over the last 30 days
3. Identify zero-invocation functions
4. Calculate cost for each:
   - Allocated memory × provisioned concurrency cost
   - Associated CloudWatch log storage
   - Any attached layers or VPC configurations
5. Generate a cleanup report:
   "Found 12 unused functions consuming an estimated $340/month.
    Recommend archiving 10 (no recent changes) and investigating
    2 (recently modified but never invoked)."

Cross-Region Resource Management

For organizations with multi-region deployments, AI assistants can correlate resources across regions using cloud MCP servers:

This multi-region capability is especially valuable for global enterprises that must maintain consistent infrastructure standards across geographic boundaries while complying with data residency requirements. Organizations operating in the European Union, for example, must ensure that certain data never leaves EU-based regions, and AI assistants with multi-region awareness can verify this compliance automatically during routine infrastructure audits.

What to Read Next

• Deploying Remote MCP Servers -- Host MCP servers in the cloud
• Enterprise Use Cases -- Secure data access patterns for organizations
• Enterprise & Specialized Servers -- CRM, ERP, and vertical industry servers
• Browse Cloud Provider Servers -- Find cloud MCP servers in our directory