Kiro IDE: install, specs, steering, hooks, skills, MCP and pricing

What is Kiro IDE?

Kiro is an agentic IDE from AWS that plans a feature as written specs before it writes code. You describe what you want, Kiro turns it into requirements, a design and a task list, and its agent then works through the tasks. The Kiro site says it is "Built and operated by AWS" and uses the tagline "Move beyond AI coding to agentic engineering."

The IDE is one of several Kiro surfaces. There is also a Kiro CLI, a web interface, a mobile app and Kiro Crew, a development workspace that keeps running across sessions. This guide focuses on the IDE and follows the Kiro docs as of 26 October 2026.

What sets Kiro apart:

• Spec-driven development. Requirements, design and tasks come first, as Markdown files you can review and commit.
• Steering files that give the agent lasting project context, including AGENTS.md.
• Agent hooks that run commands or prompts when files are saved or tools run.
• Open standards. The home page lists support for Agent Client Protocol (ACP), AGENTS.md, Skills and MCP.

How do you install Kiro IDE?

The installation page lists these systems:

Then:

1. Download the installer for your OS from kiro.dev/downloads and run it.
2. Sign in on first launch. The FAQ lists GitHub, Google, AWS Builder ID and AWS IAM Identity Center, and says "You do not need an AWS account to use Kiro."
3. Optionally import your VS Code settings and extensions. Kiro is based on Code OSS, so themes and Open VSX compatible extensions carry over.
4. Open a folder from the welcome page to start your first session.

If you also want the terminal agent, the Kiro CLI page gives one installer and a start command:

curl -fsSL https://cli.kiro.dev/install | bash
cd my-project
kiro-cli

How do specs work in Kiro?

The specs docs define specs as "structured artifacts that formalize the development process for features and bug fixes in your application." Each spec has three files:

There are two spec types. Feature specs build new capabilities and come in Requirements-First and Design-First variants. Bugfix specs diagnose and fix a bug while guarding against regressions. A Quick Spec option generates all three files in one pass without stopping for approval at each phase.

Feature requirements use EARS (Easy Approach to Requirements Syntax), written as:

WHEN [condition/event] THE SYSTEM SHALL [expected behavior]

Specs live under .kiro/specs/, with one folder per feature. The best practices page says "Specs are designed to be version-controlled, making them easily shareable across your team," so commit them next to the code they describe. If you like this way of working in other agents, the spec-driven development skill brings a similar flow to Claude Code and Codex.

How do steering files and AGENTS.md work in Kiro?

Steering files are Kiro's rules: Markdown that gives the agent lasting knowledge of your project. The steering docs list two locations:

• Workspace: .kiro/steering/ in the project root
• Global: ~/.kiro/steering/, for every workspace

When they conflict, workspace instructions win. Kiro can generate three foundation files for you: product.md (purpose, users, goals), tech.md (frameworks, tools, constraints) and structure.md (file layout, naming, architecture).

Each steering file picks an inclusion mode in its front matter:

---
inclusion: fileMatch
fileMatchPattern: "components/**/*.tsx"
---

The docs note that in Kiro CLI V1 and V2, only always files load automatically.

Kiro also reads AGENTS.md. Put it in the workspace root or ~/.kiro/steering/, and Kiro picks it up, including AGENTS.md files in subdirectories. AGENTS.md files do not support inclusion modes and are always included, so a repo set up for other agents works in Kiro without changes.

How do agent hooks work?

The hooks docs put it this way: "Hooks run shell commands or agent prompts automatically when specific events happen in your session." A command action runs in the project root and receives session context as JSON on standard input. An agent action injects a prompt into the conversation.

Triggers mentioned in the docs include PostFileSave (after a file is saved, optionally filtered by a path regex), PreToolUse (before a tool runs, so it can block it), PostToolUse and Stop, and the hook types page lists the rest, such as Prompt Submit, Session Start and triggers before and after a spec task. A typical example is a PostFileSave hook with the matcher \.(ts|tsx)$ that runs npx eslint --fix.

Hooks are JSON files in .kiro/hooks/ in the workspace, and the CLI V3 also reads ~/.kiro/hooks/ for global hooks. They activate when a session starts.

How do you add Agent Skills and MCP servers to Kiro?

Skills. Kiro supports the Agent Skills standard. The skills docs list .kiro/skills/ for a workspace (IDE, CLI and web) and ~/.kiro/skills/ globally (IDE and CLI only). A workspace skill wins over a global one with the same name.

The usual rules apply: name must match the folder and use lowercase letters, numbers and hyphens up to 64 characters, and description can be up to 1,024 characters. Kiro loads only names and descriptions at startup and pulls in the full SKILL.md when a request matches. Type /skill-name to run one directly. In the IDE, open Agent Steering & Skills in the Kiro panel, click + and choose Import a skill to copy one from a GitHub subfolder or a local folder.

Kiro also has powers, which bundle skills with an optional mcp.json behind a plugin.json manifest, installable from a marketplace or GitHub.

MCP. The MCP configuration page uses .kiro/settings/mcp.json for a workspace and ~/.kiro/settings/mcp.json for your user. Both are merged, and workspace settings win on name clashes.

{
  "mcpServers": {
    "local-server": {
      "command": "npx",
      "args": ["-y", "some-mcp-server"],
      "env": { "API_KEY": "${API_KEY}" },
      "disabled": false,
      "autoApprove": []
    },
    "remote-server": {
      "url": "https://example.com/mcp",
      "headers": { "Authorization": "Bearer ${TOKEN}" }
    }
  }
}

Local servers use command, args and env. Remote servers use url and optional headers, over HTTPS (or HTTP for localhost). Keep autoApprove limited to tools you trust. The MCP servers roundup has servers worth trying.

How much does Kiro cost?

Kiro uses credits, and the home page says there are no daily or weekly rate limits. As of October 2026, Kiro's pricing page lists:

Extra credits on paid plans cost $0.04 each, and unused monthly credits do not roll over. The Free tier covers open weight models and Claude Sonnet 4.5 with rate limits, while paid plans add premium models. By default prompts go to Auto, which the FAQ describes as "an agent that uses a mix of different frontier models," and you can also pick Claude, GPT or open weight models such as DeepSeek, MiniMax, GLM and Qwen.

On data use, the FAQ says Kiro may use some content from Free tier users and from subscribers who sign in with social login or AWS Builder ID to improve the service, with an opt-out, but not content from paid users who sign in through IAM Identity Center or an external identity provider.

How does Kiro compare with Claude Code and Codex?

Claude Code's memory docs say it reads AGENTS.md when there is no CLAUDE.md, and Codex reads AGENTS.md by default, so one AGENTS.md can steer all three. Skills need to be copied or linked into each tool's folder. For a deeper look at the other two, see Codex vs Claude Code and the Claude Code plan mode guide.

Kiro suits teams that want written requirements and designs reviewed before code, inside a familiar VS Code style editor. Claude Code and Codex suit people who prefer to plan in conversation and drive the agent from a terminal or their existing editor.

How do you reuse your skills across Kiro and other agents?

Because Kiro uses standard SKILL.md folders, the same skill can run in Kiro, Claude Code and Codex. Kipt keeps your prompts and skills as Markdown in a private GitHub repo you own, imports skills from a GitHub link, a folder or a .zip, and lets you search them with Cmd+K. From there you can import a skill into Kiro with Import a skill or copy the folder into .kiro/skills/. To turn a prompt into a skill, try the free SKILL.md generator.