MCP Servers

Model Context Protocol (MCP) is an open protocol that standardises how applications provide context to LLMs. With MCP, Kiro can connect to various data sources and tools through MCP servers, extending its capabilities beyond built-in features. This allows Kiro to interact with databases, APIs, file systems, and other services in a standardised way.

AWS MCP Servers

AWS provides 50+ MCP servers for interacting with AWS services. Find the full catalog at awslabs.github.io/mcp.

Recommended Servers by Use Case

Documentation

Essential for any AWS development:

Server	Purpose
AWS Documentation	Get latest AWS docs and API references
AWS Knowledge	Documentation, code samples, official content

Infrastructure

For deploying and managing AWS resources:

Server	Purpose
AWS CDK	CDK development with security compliance
CloudFormation	Direct CloudFormation resource management
Terraform	Terraform workflows with security scanning
Cloud Control API	Full CRUDL operations with security scanning
EKS	Kubernetes cluster management
ECS	Container orchestration and deployment
Serverless (SAM)	Serverless app lifecycle with SAM CLI

Databases

For data storage and queries:

Server	Purpose
DynamoDB	Complete DynamoDB operations
Aurora PostgreSQL	PostgreSQL via RDS Data API
Aurora MySQL	MySQL via RDS Data API
DocumentDB	MongoDB-compatible operations
Neptune	Graph database queries
Redshift	Data warehouse queries

AI/ML

For AI and machine learning:

Server	Purpose
Bedrock Knowledge Bases	Query knowledge bases with citations
Nova Canvas	AI image generation
SageMaker	Model development and deployment
Bedrock Data Automation	Analyse documents, images, video, audio

DevOps

For monitoring and operations:

Server	Purpose
CloudWatch	Metrics, alarms, and logs analysis
CloudTrail	API activity and audit logs
Cost Explorer	Detailed cost analysis
Pricing	Pre-deployment cost estimation
AWS Diagram	Generate architecture diagrams

Highlight: AWS Diagram Server

The AWS Diagram MCP Server can automatically generate architecture diagrams from your infrastructure code or descriptions. Great for documentation and presentations!

Configuring MCP Servers in Kiro

1. Add MCP server configuration

   To define MCP servers, add them as you normally would in any MCP client. The configuration follows the standard MCP format.

   

2. Example configuration

   Here’s an example MCP server configuration for AWS Documentation:

   {
     "mcpServers": {
       "aws-docs": {
         "command": "uvx",
         "args": ["awslabs.aws-documentation-mcp-server@latest"],
         "env": {
           "AWS_PROFILE": "default"
         }
       }
     }
   }

3. Restart Kiro

   After adding configuration, restart Kiro to load the new MCP servers.

MCP vs Powers

For common use cases, consider using Kiro Powers instead:

Feature	MCP Servers	Powers
Setup	Manual JSON config	One-click install
Includes	Tools only	Tools + Steering + Hooks
Flexibility	Full control	Pre-configured
Best for	Custom setups	Common workflows

Note

Powers are built on MCP servers. If a Power exists for your use case, it’s usually easier to install the Power. Use direct MCP configuration when you need custom setups or servers not available as Powers.

Practical Example: Cost Estimation

With the AWS Pricing MCP Server configured, you can estimate costs before deploying:

"Estimate the monthly cost for running 3 t3.medium EC2 instances
with 100GB EBS storage each in ap-southeast-2"

Kiro will query current AWS pricing and provide a detailed cost breakdown.

Further Resources

• AWS MCP Catalog: awslabs.github.io/mcp
• MCP GitHub: github.com/awslabs/mcp
• Community servers: promptz.dev hosts additional MCP configurations

With Steering Docs, Agent Hooks, Powers, and MCP Servers configured, Kiro is now equipped with the tools and knowledge it needs. Let’s start executing the build tasks.