A suite of MCP servers that provide AWS documentation, deployment guidance, and cloud-native tooling through the Model Context Protocol.
Configuration
View docs{
"mcpServers": {
"awslabs-mcp": {
"url": "https://knowledge-mcp.global.api.aws",
"headers": {
"FASTMCP_LOG_LEVEL": "ERROR"
}
}
}
}You can connect your language models to AWS data, services, and documentation through a suite of Model Context Protocol (MCP) servers. These servers expose AWS-specific capabilities as MCP endpoints, enabling AI tools and IDEs to query up-to-date AWS docs, perform infrastructure tasks, manage resources, and automate workflows with context-rich guidance.
To use an MCP server, you configure an MCP client (for example, Kiro, Cursor, Windsurf, or an IDE plugin) to connect to the MCP endpoint and load the available tools. Each server provides access to AWS documentation, infrastructure guidance, data operations, or deployment workflows. You typically choose a server from the available options, add it to your MCP client, and then invoke its tools as part of your AI-assisted development tasks. When you issue a task, the MCP client will route queries to the server, which returns the contextual guidance, documentation references, code examples, or operational steps you need.
Prerequisites you need before installing any MCP server:
- Install Python and the uv command runtime from Astral if your flow requires it.
- Ensure you have AWS credentials configured with access to the services you plan to use.
- Have an MCP client ready (for example, Kiro, Cursor, or VS Code) to load the server. The following steps show a concrete flow to set up an MCP server for your development environment.
# Step 1: Install the MCP runtime tool if required (example placeholder)
# Your environment may already include uv or an equivalent runtime
# Install uvx tooling per your setup if needed
# Step 2: Install a server you want to use via an MCP client config
# Example: AWS Knowledge MCP Server (http) and AWS Documentation MCP Server (stdio)
# AWS Knowledge MCP Server (HTTP) installation via an MCP client requires a URL
# This is an example URL pattern; adapt to your actual installation flow
# Step 3: Add a local or remote MCP server entry to your MCP client config
# The following JSON snippet demonstrates the configuration structure you would place
# in your MCP client settings. Replace placeholders with your actual values.{
"mcpServers": {
"aws-knowledge-mcp": {
"type": "http",
"url": "https://knowledge-mcp.global.api.aws"
}
}
}If you prefer a local, self-contained setup, you can run a stdio MCP server from your development machine or container. The readme provides a large catalog of concrete server configurations. Choose the server that fits your workflow, and configure the client accordingly.
- Some MCP servers are designed for remote hosting and provide an HTTP URL. Others run locally as stdio servers and require a command and arguments to start. Always follow the exact command and argument structure shown in the specific server’s setup snippet.
- If a server requires environment variables, include those in your MCP client configuration and, where applicable, in your local runtime environment. For example, a server might specify FASTMCP_LOG_LEVEL=ERROR to reduce verbosity.
- When a server is provided as a fully managed AWS service, you can connect to it directly via the MCP client without managing the hosting infrastructure.
Plan and implement security controls for MCP usage. Use IAM-based permissions for cloud actions, enable auditing, and ensure credentials and tokens are managed securely. Apply standard software security practices to avoid exposing credentials within MCP configurations.
If you encounter issues, verify that each MCP server entry uses a valid URL (for http servers) or a valid command and argument list (for stdio servers). Check any required environment variables and confirm that the MCP client is configured to allow the server to be reached from your development environment. If a server’s startup sequence includes building steps, ensure those complete before starting the server.
Sample server entries illustrate how to connect to AWS Knowledge MCP Server (HTTP) and AWS Documentation MCP Server (stdio). You would adapt these samples to your own MCP client configuration and environment.
Provides general AWS API interactions with safety checks and validated calls.
Gives access to AWS documentation and references through MCP.
Remote knowledge base for latest AWS docs and references.