Kiro

Bring engineering rigor to agentic development

Coding Freemium Free: 50 credits/month with open weight models and Claude Sonnet 4.5; Pro: $20/month 1,000 credits; Pro+: $40/month 2,000 credits; Pro Max: $100/month 5,000 credits; Power: $200/month 10,000 credits. Add-on credit packs available with 12-month expiry. ★ 4.5 Visit Website →

About

Kiro is an AI-powered IDE, CLI, and web interface from Amazon. It uses spec-driven development, Agent Hooks, and advanced steering to help developers go from vibe coding to viable code. Supports multiple frontier models (Claude Opus 5/4.8, Sonnet 5, OpenAI GPT-5.6, Qwen, DeepSeek, MiniMax M2.5), native MCP integration, and VS Code plugin compatibility. Reached GA in November 2025 with autonomous agent capabilities, multi-repo context, asynchronous task execution, and native integrations with Jira, GitHub, and Slack. CLI V3 early access introduces Introspect subagent, global hooks, and persistent model preferences. IDE 1.0 brings new permissions system, structured hooks format, and session export.

Key Features

✓Spec-driven development (requirements→architecture→tasks→implementation, Quick Plan mode)
✓Autonomous agent (persistent context, up to 10 concurrent async tasks)
✓Agent Hooks automation (file write triggers, global hooks, v1 JSON format)
✓Steering rules customization (Markdown format manual steering files)
✓Native MCP support (automatic OAuth, network failure recovery, lazy auth)
✓CLI V3 early access (Introspect subagent, global hooks, persistent model preferences, /upgrade-agent migration)
✓Multi-model support (Claude Opus 5/4.8, Sonnet 5, OpenAI GPT-5.6 Sol/Terra/Luna, MiniMax M2.5, Qwen, DeepSeek)
✓Multi-repo context understanding
✓Native integrations with Jira, GitHub, Slack
✓Enterprise SSO (Okta, Microsoft Entra ID, AWS IAM Identity Center)
✓Custom subagents and skill packages (Markdown-based agent configs)
✓Web automations (scheduled tasks, automatic PRs)
✓Session auto-restore, thinking display, session export (ZIP format)
✓IDE 1.0 permissions system (capability-based permissions model)
✓Parallel task execution, requirements analysis (automated reasoning for logical inconsistencies)
✓Adjustable reasoning depth (--effort flag: low/medium/high/xhigh/max)
✓Display and accessibility settings (streaming animations, ASCII art, icon toggles)
✓AWS GovCloud (US) support (user activity monitoring, usage dashboard, daily CSV telemetry, optional prompt logging)

Who it's for

Team developers maintaining long-lived codebases who need design decisions on the record
Engineers already running Jira / GitHub / Slack workflows who want an agent wired into them
Not for casual users who just want a throwaway script — the spec overhead does not pay off

Pros & cons

Pros

Spec-driven flow persists requirements and architecture as files in the repo, so rework has a paper trail instead of relying on chat history
One subscription covers the IDE, CLI and web surfaces — no per-surface billing
Native MCP plus Jira / GitHub / Slack integrations remove the need for custom glue
Steering rules and specs are Markdown inside the repo, so they go through code review and version control instead of living in editor-local settings

Limitations

Credit-based pricing ramps up fast under heavy use (especially at high --effort); the free tier is only 50 credits/month
Desktop (Windows / macOS / Linux) and web only — no mobile client
CLI V3's Introspect subagent and global hooks are still early access, so those interfaces may change
Running the full spec flow for a one-line change is slower; you have to judge when to bypass it