AWS Enables Remote Kiro and VS Code Access to SageMaker

An AWS Big Data blog post published June 2026 documents AI-assisted data development using Kiro and Amazon SageMaker Unified Studio. The post covers remote IDE connectivity - a feature that launched for VS Code in September 2025, and for Kiro and Cursor in March 2026 - and describes newer AI-assistant integrations including steering files and MCP server configurations. According to the blog, the AWS Toolkit for Visual Studio Code can connect local IDEs - specifically Kiro, VS Code, and Cursor - to a SageMaker Unified Studio Space using a secure remote access connection. The blog states the connection provides the IDE with access to the Space's file system, compute, and data services without additional SSH key management. The AWS documentation page on remote access describes the same model and lists deep link, AWS Toolkit authentication, and SSH terminal as supported connection methods.

Per the AWS blog, two features are central to the AI-assisted workflow. The blog reports that when connecting Kiro to SageMaker Unified Studio, Kiro generates AGENTS.md and smus-context.md files that provide the AI assistant with context about the project environment, including configuration, environment details, and utilities for discovering the data catalog and project structure. The post also documents MCP server discovery and configuration, naming servers such as smus_local and aws-dataprocessing, which the blog says provide programmatic access to an AWS Glue Data Catalog, Amazon Athena queries, and SageMaker Unified Studio metadata from a connected agent. The AWS remote-access documentation explains the connection as a secure tunnel that preserves the Studio security boundary while enabling local IDE extensions and tooling to operate over the remote session.

Industry pattern: cloud-hosted development workspaces increasingly expose remote IDE hooks so practitioners can keep local editor workflows while running code and queries on managed cloud compute. Integrations that attach AI assistants to a cloud project context - via steering files and MCP-style context servers - reduce the friction of transferring project metadata and data discovery steps into prompt-based workflows. For teams that rely on managed catalogs like AWS Glue and interactive query engines like Amazon Athena, exposing catalog and query metadata through MCP endpoints aligns with broader efforts to make data catalogs actionable inside editor and assistant UIs.

The AWS blog and docs provide the actionable path: install the AWS Toolkit in the local editor, authenticate via the Toolkit or deep link, and open a SageMaker Space to establish the remote connection. Independent tutorials (classmethod) demonstrate required IAM and instance configuration steps, and AWS Skill Builder offers a 30-minute course to learn the workflow. AWS notes not all local IDE extensions work over remote sessions. West Loop reports the feature is available in all Regions where SageMaker Unified Studio is available.