Provides access to AWS documentation, guidance, and workflows through MCP for enhanced AI-assisted cloud development.
Configuration
View docs{
"mcpServers": {
"awslabs-mcp": {
"url": "https://knowledge-mcp.global.api.aws",
"headers": {
"FASTMCP_LOG_LEVEL": "ERROR"
}
}
}
}You can extend AI capabilities by connecting your language models to AWS data and tools through MCP servers. These servers expose AWS documentation, guidance, and workflows as programmable endpoints that your MCP clients can use to enrich responses, automate tasks, and enable cloud-native development with safety and structure.
You use MCP servers by configuring an MCP client to connect to one or more MCP servers. Each server acts as a provider of AWS-specific knowledge, data access, and tooling. Include the servers you need in your MCP client configuration, then prompt your AI agent to call the servers as part of its workflow. For example, you might enable an AWS Documentation MCP Server to fetch up-to-date API references while your agent designs infrastructure, or enable a CloudFormation/CDK related server to validate resources and generate IaC templates.
Follow these practical steps to set up MCP servers for AWS. Ensure you have a development environment with basic tooling and AWS credentials. The workflow below covers common prerequisites and then demonstrates adding the two known MCP server configurations described in this guide.
Prerequisites
- Install Python and a runtime you will use for MCP tooling (examples show uv and related scripts).
- Install Node.js or Python tooling if you plan to run local tooling that depends on those runtimes.
- Configure AWS credentials with access to the services you plan to use.
Step 1: Create the MCP server entries
- HTTP MCP server (remote) for official AWS knowledge sources
- STDIO MCP server (local) that runs on your machine
Step 2: Start and test the MCP clients
- Ensure your MCP client is aware of the servers you added and can reach them via the configured transports.The following MCP server entries are examples you can start with. They show how to connect to a remote AWS Knowledge MCP Server via HTTP and how to run a local core MCP server via standard input/output.
// HTTP: AWS Knowledge MCP Server (remote)
{
"type": "http",
"name": "aws_knowledge",
"url": "https://knowledge-mcp.global.api.aws",
"args": []
}
// STDIO: Core MCP Server (local)
{
"type": "stdio",
"name": "core_mcp",
"command": "uvx",
"args": ["awslabs.core-mcp-server@latest"],
"env": {
"FASTMCP_LOG_LEVEL": "ERROR"
}
}
"],Two example MCP server configurations are provided to illustrate usable MCP entries. Use these exact snippets in your MCP client configuration if you want to mirror the setup described here.