Kiro is an AI-powered development environment that helps you build software from prototype to production. One unified agent harness powers every surface (IDE, CLI, Web, and Mobile) so your configuration, specs, and steering work everywhere.
Pick the surface that fits your workflow. Your .kiro/ configuration is shared across all of them.
| I want to... | Use |
|---|---|
| Plan a feature with requirements, design, and tasks | Specs |
| Fix a bug with root cause analysis and regression prevention | Bugfix Specs |
| Chat and build iteratively without a plan | Chat (IDE) or start a session (CLI) |
| Enforce project standards automatically | Steering |
| Automate actions on file save, tool use, or task completion | Hooks |
| Connect external tools and APIs | MCP |
| Control what the agent can access | Permissions |
| Create specialized agents for specific workflows | Custom agents |
| Extend the agent with reusable instruction packages | Skills |
| Add tools with built-in knowledge that activate on demand | Powers |
| Delegate parallel tasks to focused sub-agents | Sub-agents |
| Undo agent changes or fork a conversation | Checkpoints and rewind |
| Keep secrets and sensitive files away from the agent | Kiroignore |
| Keep long sessions productive within context limits | Compaction |
Every Kiro surface is a front end to the same unified agent harness. Your steering files, permissions, hooks, MCP servers, and custom agents work identically whether you're in the IDE, CLI, or Web — start a spec in the IDE, continue it from the CLI, hand off implementation to the Web agent, and everything stays in sync through .kiro/.
Each surface adds its own way of working on top: the IDE brings editor integration, the CLI brings terminal-native and headless workflows, and the Web brings zero-setup sandboxed sessions. How Kiro works explains the architecture, the agent loop, and where every capability plugs in.
Kiro Documentation