Kiro IDE vs. CLI

Two surfaces, one agent harness: choose by interface

Overview

Kiro ships as a full-featured IDE (desktop, graphical) and a CLI (terminal). Since Kiro CLI v3, the CLI runs on the same unified agent harness as the IDE and Web, so the main difference is the interface, not the feature set.

What changed with CLI v3

Older comparisons said "Specs, Powers, and autonomous mode are IDE-only." That is no longer true. With v3 the CLI brings the shared agent harness to the terminal, so spec mode, Powers management, and autonomous mode with a local sandbox are available there too. This page is updated to that state.

Feature Comparison

Feature IDE CLI
Spec-driven development✓✓ (/spec new, since v3)
Powers✓✓ (manage in a local v3 session)
Autonomous mode & local sandbox✓✓ (/autonomous on, GA in IDE & CLI)
Hooks & Steering✓✓ (standalone + global hooks)
MCP Servers✓✓
Custom Agents✓✓ (universal config, /upgrade-agent)
Capability-based permissions✓✓ (deny > ask > allow)
Shared .kiro/ config✓✓
Headless mode (CI/CD)—✓
API key auth (headless)—✓
Cloud sessions (--cloud)—✓ (us-east-1)
Tangent mode (/tangent)—✓
Visual debugging / IntelliSense✓—
Drag & drop (images/PDFs)✓—
VS Code dev containers— (not currently, MS extension not on Open VSX)✓ (CLI runs in dev containers)

When to Use Which

Choose IDE when…

  • You need a full editor with visual debugging and IntelliSense
  • You prefer the graphical source-control panel and file explorer
  • You like editing specs visually in the spec explorer
  • You pass images or PDFs as context via drag & drop
  • You use Open VSX extensions

Choose CLI when…

  • You automate tasks in CI/CD pipelines
  • You run headless (no GUI available)
  • You want cloud sessions that keep running after you disconnect
  • You delegate multi-step work to workflows
  • You work on a remote server over SSH
No longer decisive

Specs, Powers, and autonomous mode are no longer reasons to pick the IDE by necessity. Since v3 they exist in the CLI too. The choice today mainly comes down to whether you want a graphical interface or work in the terminal.

Frequently Asked Questions

What is the difference between Kiro IDE and Kiro CLI?

Kiro IDE is a full desktop IDE with a graphical interface. Kiro CLI is a terminal-based AI tool for interactive work, headless automation, and CI/CD. Since CLI v3 both run on the same unified agent harness, so they share Specs, Hooks, Steering, MCP, Powers, and permissions. The main difference is the interface, not the feature set.

Can you use Kiro IDE and CLI simultaneously?

Yes. IDE and CLI use the same account, models, and .kiro/ configuration. You can use the IDE for feature development and the CLI for CI/CD automation in parallel.

Which features stay largely IDE-specific?

Visual debugging, editor IntelliSense, the graphical source-control panel, and drag-and-drop for images/PDFs. Specs, Powers, and autonomous mode with a local sandbox are available in the CLI too since v3.

What is typical for the CLI?

Headless mode for CI/CD (API key auth), tangent mode (/tangent), cloud sessions (kiro-cli --cloud), workflows, and a terminal UI with session management.

Dev containers: a limit of the IDE

The Kiro IDE currently cannot work inside VS Code dev containers

The well-known Dev Containers extension (ms-vscode-remote.remote-containers) is a Microsoft-proprietary extension and is not available on the Open VSX Registry, where the Kiro IDE gets its extensions. As a result, you currently cannot open a project in a dev container from the Kiro IDE the way VS Code does. It is an open, repeatedly requested feature in the Kiro bug tracker.

Important for choosing between IDE and CLI: the Kiro CLI does run inside dev containers. If you rely on containerized, isolated development environments (e.g. for security when running AI agents), you can use the CLI in the container today, while IDE support is still pending.

Sources: