Read this on the main serverless docs site

Serverless MCP Server

The Serverless MCP Server is an implementation of the Model Context Protocol (MCP) that provides a set of tools for working with cloud resources and services across multiple Infrastructure as Code (IaC) solutions. It enables AI assistants to interact with your cloud infrastructure, helping you diagnose issues, retrieve information, and understand your serverless applications.

::youtube{id="FW6IpZv_xUU"}

Features

Multiple Transport Options: Run the server using HTTP/SSE or stdio transport
Comprehensive Cloud Resource Information: Detailed metrics, configurations, and logs for serverless functions, roles, queues, storage, and API resources
Infrastructure as Code Integration: List and analyze resources defined in multiple IaC solutions including Serverless Framework, AWS CloudFormation, and SAM templates
Parallel Data Fetching: Efficient retrieval of metrics and logs across multiple resources
Error Grouping and Analysis: Intelligent aggregation of error logs for easier troubleshooting
Flexible Time Range Support: Customizable time periods for metrics and logs analysis
Service Summary Tool: Consolidated view of multiple resource types in a single request

Getting Started

Install the Serverless Framework if you haven't already:

bash

npm install -g serverless

Then, you configure the MCP Server in your host application:

json

{
  "mcpServers": {
    "serverless": {
      "command": "serverless",
      "args": ["mcp"]
    }
  }
}

For detailed setup instructions, see the Setup Guide.

Example Prompts

Provide a complete overview of my serverless application's resource relationships and data flows based on cloud data
Display the deployment history of my serverless application for the past 7 days
Identify which functions of my application have experienced the most errors
Analyze my CloudWatch Alarms history and highlight recurring issues
Detect unusual traffic patterns in my API Gateway during the past month
Examine my SQS queues and determine if any messages are stuck

Best Practices

Always specify the profile explicitly when using MCP tools, especially in environments with multiple AWS accounts or roles
If the region is not specified in your service configuration, always provide the region parameter
For Serverless Framework projects, check the provider.region setting in your infrastructure configuration file
Use AWS SSO for enhanced security in enterprise environments
Regularly rotate your access keys if using long-term credentials

Next Steps

Available Tools: Explore the tools provided by the MCP Server
AWS Integration: Learn how to configure AWS credentials
Setup: Learn how to set up the MCP Server in your AI assistant

README