AWS EC2 Instance Connect MCP Server

When this API is exposed as a set of tools to an AI coding assistant through the Model Context Protocol (MCP), it unlocks significant value for automating and securing infrastructure workflows. An AI agent, operating within a developer's IDE or a chat interface, gains the ability to directly orchestrate secure instance access as part of higher-level tasks. For example, instead of the developer manually generating a temporary key pair and calling the AWS CLI, they can instruct the AI to "securely connect to the production web server instance i-1234567890abcdef0 to check the application logs." The AI, via the MCP server, can then execute the workflow: generate a fresh, one-time SSH key pair locally, use the SendSSHPublicKey tool to push the public key to the specified instance, and provide the temporary private key or connection command back to the user or an integrated SSH client. This integration transforms instance access from a manual, error-prone step into a secure, tool-driven operation embedded within natural language instructions.

Practical workflow examples demonstrate the dynamic tasks an AI agent can perform. A developer could instruct: "Help me deploy a configuration change to all instances tagged with 'Environment: staging'." The AI agent could, in sequence, query AWS to list the target instances, then use the EC2 Instance Connect API to establish a secure, temporary connection to each instance one by one, execute the necessary configuration management commands (like applying a new config file via SCP or running a script), and report back on the success or failure for each instance. Another example is automated diagnostics: "The monitoring system is alerting on high CPU on instance i-... . Connect to it and gather a snapshot of running processes and network connections." The AI would initiate a temporary connection, run diagnostic commands like top -bn1 and ss -tuln, and return the structured output for analysis, all without the developer needing to handle or store any sensitive credentials on their local machine.

Crucially, while the API itself does not require direct authentication at the endpoint level (as the SendSSHPublicKey action is authenticated via the caller's IAM permissions), a robust security posture is mandatory. The IAM entity (user, role, or assumed role) invoking the API must have explicit permissions for the ec2-instance-connect:SendSSHPublicKey action, scoped to the specific instance IDs or tags using resource constraints, adhering to the principle of least privilege. Developers must ensure the EC2 instances are configured with the required IAM role (EC2InstanceConnect) or have the EC2 Instance Connect agent installed and configured. Best practices include enforcing tag-based policies for access control, logging all API calls via AWS CloudTrail for audit purposes, and integrating this tool-based access with broader identity governance systems. The temporary nature of the keys is a primary security feature, but it must be complemented by proper network controls (like Security Groups) and instance-level hardening to ensure a comprehensive defense-in-depth strategy.