AWS Kiro CLI 2.1: Real-Time Shell Streaming, Device Auth, and RHEL Support

AWS Kiro shipped CLI version 2.1.0 on April 24 with a set of practical improvements aimed at enterprise and server environments. None of these are headline features, but they address real friction points that have made Kiro awkward in certain workflows.

Shell Output Streams as It Runs

The most immediately useful change is real-time shell output streaming. Previously, when Kiro ran a shell command, it buffered the output until the process completed before showing you anything. Now it streams line by line as the process runs.

For anything that takes more than a few seconds, this is a material quality-of-life improvement. Build processes, test runs, package installs, and long scripts all give you live feedback instead of a blank wait followed by a wall of text.

MCP Tools Load on Demand

Kiro now loads MCP tool definitions on demand rather than sending all definitions upfront when a session starts. If you have multiple MCP servers configured, this keeps the context window from filling up with tool definitions for capabilities you’re not using in a given session.

The practical effect is more context space for actual work, particularly if you’ve accumulated a large set of MCP tools over time.

Skills as Slash Commands

Reusable agent instructions stored in .kiro/skills/ directories are now directly invokable as /skill-name commands. Previously you had to navigate to the skill or reference it by path. The slash command syntax makes it considerably faster to invoke a frequently-used skill mid-session without breaking flow.

Device Flow Authentication

Kiro 2.1 adds device flow authentication for signing in from SSH sessions and containers. You get a URL and a one-time code; enter the code on any browser to authenticate. No port forwarding required.

This closes a gap that made Kiro awkward in common developer environments: remote machines accessed via SSH, cloud dev environments, and containers running without browser access. Device flow is the standard approach for CLI auth in headless environments (GitHub CLI uses the same pattern), and the absence of it in earlier Kiro releases was a real obstacle for anyone trying to use it in CI/CD or server-side workflows.

RHEL Support

The Kiro terminal UI now works on Red Hat Enterprise Linux, matching feature parity with macOS, Windows, and other Linux distributions. For developers working in enterprise environments where RHEL is standard, this removes a blocker.

Context

Kiro launched in July 2025 as Amazon’s answer to Cursor: a spec-driven agentic IDE built on VS Code, powered by Claude Sonnet. It’s differentiated by its “steering files” approach, where you define project requirements, design docs, and agent behaviors in structured files that the agent can consult and update. The CLI mode (separate from the IDE) targets developers who prefer terminal-first workflows or need to run Kiro in headless environments.

CLI 2.0, released April 13, added Windows 11 support and a headless CI/CD mode. Version 2.1 continues pushing the CLI toward enterprise viability, with the device auth and RHEL additions specifically addressing gaps that would block adoption in regulated or security-conscious organizations.

Source: Kiro CLI Changelog