Kiro
Agent-centric IDE with spec-driven development by Amazon
Install with Homebrew
brew install --cask kiroQuick Take: Kiro
By August 2026 Kiro is no longer a vague preview: it is a credit-priced, multi-tier AWS AI IDE with a clear Free→Power ladder, Pro Max at $100, and premium Claude 4.x-class models. Spec-driven development and hooks remain the product thesis, more process, more predictability. Credit burn is the main risk. For teams that value structured agent work and AWS governance, Kiro belongs on the shortlist beside Cursor and CLI agents.
Best For
- •Teams that want spec-traceable AI code changes
- •AWS-heavy product and platform engineers
- •Orgs needing SSO, analytics, and seat-level credit governance
- •Developers willing to plan before prompting
What is Kiro?
Kiro is Amazon Web Services' agent-centric IDE for spec-driven, agentic software engineering. Unlike free-form AI editors that jump straight from a prompt to a patch, Kiro pushes you to capture requirements, design decisions, and acceptance criteria first, then agents implement and verify against that spec. The product reached general availability after its 2025 preview and, as of August 2026, is a full commercial offering with Free through Power individual tiers, matching team plans, CLI/web surfaces, and enterprise SSO via AWS IAM Identity Center. Kiro is built on VS Code (Code OSS) foundations, so extensions, themes, and keybindings transfer with little friction. Its differentiators are the planning layer (specs and task breakdown), Steering Hooks / agent hooks for event-driven quality automation, multi-model routing (Auto plus selectable Claude and open-weight models), and deep AWS-oriented workflows. Free-tier users get 50 credits per month with open-weight models such as Qwen3 Coder Next, DeepSeek v3.2, and MiniMax 2.1, plus Claude Sonnet 4.5 access subject to rate limits. Paid plans unlock premium models including Auto, Claude Sonnet 4.6, and Claude Opus 4.8, with transparent overages at $0.04 per credit. Pricing (individual, per user/month as published on kiro.dev/pricing in August 2026): Free $0 (50 credits); Pro $20 (1,000 credits); Pro+ $40 (2,000); Pro Max $100 (5,000); Power $200 (10,000). Team plans mirror those credit envelopes and add consolidated billing, usage analytics, SAML/SCIM SSO, and org controls. First paid upgrade with social login or AWS Builder ID gets a $20 subscription credit. Unused monthly credits do not roll over; individual add-on credit packs do roll for up to 12 months. Kiro is actively shipping: mid-2026 coverage includes iOS/remote session positioning, RedMonk analysis of its cloud-session thesis, and heavy community discussion of credit burn when running multiple agents. Competitors in the same budget band include Cursor, Claude Code inside the Anthropic ecosystem, Warp agents, Windsurf under Cognition, and newer terminal agents such as Meta Muse Code. Kiro's bet is predictability and enterprise governance over pure vibe-coding speed. On Mac, Kiro installs as a native desktop app (Apple Silicon supported) plus CLI. You do not need an AWS account for individual social/Builder ID sign-in, though AWS identity and GovCloud pricing paths exist for regulated teams (~20% higher, no Free tier). Honest take: if you want structured, reviewable agent output, especially on AWS-shaped systems, Kiro is now a shortlist product. If you want the fastest free-form edit loop, Cursor or CLI agents may still feel snappier; many teams run both.
Key Features
Spec-Driven Development
Start features by capturing requirements, design notes, and acceptance criteria. Agents implement task-by-task against the spec instead of guessing from a single prompt. Better for multi-file, multi-service work where 'almost right' is expensive.
Steering Hooks and Agent Automation
Event-driven hooks run agents on code changes, tests after edits, docs when APIs change, migration validation before apply. Turns agents into continuous quality workers, not only chat responders.
Multi-Model Access (Auto + Premium + Open Weights)
Free tier: open-weight models (Qwen3 Coder Next, DeepSeek v3.2, MiniMax 2.1) and Claude Sonnet 4.5 with rate limits. Paid: Auto routing plus premium models including Claude Sonnet 4.6 and Claude Opus 4.8 (availability varies by region). Credits scale with model choice, e.g. Sonnet 4.6 can cost more credits per task than Auto.
Credit-Based Pricing Without Daily Caps
Plans sell monthly credits with no daily/weekly rate-limit cliff on paid tiers. Individuals buy add-on packs at $0.04/credit; teams can enable overages at the same unit price. Usage updates in the IDE dashboard roughly every five minutes.
AWS-Native and Enterprise Controls
Strong fit for Lambda, API Gateway, DynamoDB, IAM, and infra-as-code workflows. Team/Enterprise paths add SAML/SCIM via AWS IAM Identity Center, org dashboards, analytics, and GovCloud options.
IDE, CLI, Web, and ACP Surfaces
One subscription covers Kiro IDE, Kiro CLI, Kiro on the web (preview on paid plans), ACP-compatible IDEs, and CI-oriented automation. Credits are shared across surfaces, no separate cloud-compute tax for web preview.
VS Code Extension Compatibility
Code OSS base means marketplace extensions, themes, and keymaps largely work. Teams migrate from VS Code/Cursor without rebuilding their entire editor stack.
Who Should Use Kiro?
1Enterprise / regulated developer
Implement a payment endpoint with explicit validation, logging, and audit requirements. Capture them as a Kiro spec, let agents implement against checklist items, and use hooks to run compliance tests after every change. Specs become the review artifact auditors want.
2AWS serverless engineer
Describe an S3 → Lambda → DynamoDB → CloudFront pipeline in a spec. Kiro scaffolds functions, IaC, IAM least-privilege stubs, and tests while you refine acceptance criteria instead of babysitting every file.
3Team lead standardizing AI coding
Roll out Pro/Pro+ seats with SSO, usage analytics, and shared overage policy so individuals do not silently hit free-tier walls. Specs and hooks encode team norms (tests, docs, security checks) instead of relying on each developer's prompt craft.
Install Kiro on Mac
Kiro ships a native macOS app and CLI. Sign in with GitHub, Google, AWS Builder ID, or IAM Identity Center, an AWS account is optional for individual use.
Download or install CLI
Get the macOS build from kiro.dev/downloads, or install the CLI via the official installer script documented on kiro.dev.
Install and sign in
Open the DMG, drag Kiro to Applications, launch, and authenticate. Import VS Code settings/extensions if prompted.
Create a first spec
Open a repo and start a structured spec for a small feature. Watch credit usage in the subscription dashboard so Free-tier 50-credit limits do not surprise you.
Pro Tips
- • Start on Free or the paid-plan trial credit path before committing to Power.
- • Prefer Auto for routine tasks; reserve Opus/Sonnet premium models for hard multi-file work to conserve credits.
- • Enable team overages only after reviewing analytics, multi-agent runs can burn credits quickly.
- • Students should check kiro.dev/students for current education offers.
Configuration Tips
Instrument hooks before large refactors
Wire test/lint/typecheck hooks first so agents cannot merge broken intermediate states while executing long specs.
Template common specs
Save templates for API endpoints, React components, migrations, and Lambdas so planning stays consistent across the team.
Budget credits by model
Track that premium models cost more credits per task than Auto. Heavy Opus use belongs on Pro Max/Power, not Free.
Use CLI for headless loops
Kiro CLI fits CI review bots and unattended tasks; keep interactive IDE sessions for spec authoring and hard debugging.
Alternatives to Kiro
Kiro competes in the 2026 AI IDE / coding-agent market against free-form editors and CLI agents.
Cursor
Cursor emphasizes free-form Agent/Tab speed and Composer models. Pro sits near $20/mo like Kiro Pro, but usage pools and soft limits differ from Kiro's explicit credits. Pick Cursor for rapid edit loops; Kiro for spec-first, enterprise-shaped workflows.
Claude Code / ChatGPT Codex
Anthropic and OpenAI CLI/desktop agents excel at long unattended runs. They lack Kiro's built-in spec product surface and AWS team controls but may win on raw agent endurance.
Windsurf (Cognition)
Agentic IDE lineage under Cognition after 2025 ownership changes. Strong agent UX; different model/billing story. Compare if you want Cascade-style flows without AWS-centric packaging.
Warp
Terminal-native agent with credit pricing (Build/Max/Business). Better when your work is shell-first rather than full IDE.
Meta Muse Code
New August 2026 terminal coding agent (Muse Spark). Watch if you want a lighter CLI rival; less mature enterprise story than Kiro.
VS Code + GitHub Copilot
Familiar editor with Copilot models (including GPT-5.6 family in 2026). Less opinionated about specs; lower switching cost if you already live in Copilot.
Pricing
Verified against kiro.dev/pricing (August 2026). Individual - Free, $0/mo: 50 credits; open-weight models (Qwen3 Coder Next, DeepSeek v3.2, MiniMax 2.1) and Claude Sonnet 4.5 with rate limits. - Pro, $20/user/mo: 1,000 credits; premium models; add-on credits $0.04 each. - Pro+, $40/user/mo: 2,000 credits; premium models; add-ons $0.04. - Pro Max, $100/user/mo: 5,000 credits; premium models; add-ons $0.04. - Power, $200/user/mo: 10,000 credits; premium models; add-ons $0.04. Team plans mirror Pro/Pro+/Pro Max/Power credit tiers with consolidated billing, usage analytics, SAML/SCIM via AWS IAM Identity Center, org dashboard, and enterprise security controls. Team overages $0.04/credit when enabled. Notes: First paid upgrade via social login or AWS Builder ID includes a $20 subscription credit. Monthly plan credits do not roll over; purchased individual add-on packs roll up to 12 months. Taxes extra; GovCloud ~20% higher, no Free tier. Premium model list includes Auto, Claude Sonnet 4.6, Claude Opus 4.8 (not all models in all regions).
Pros
- ✓Spec-first workflow produces more reviewable, requirement-traceable agent output
- ✓Transparent credit pricing with Pro Max mid-tier and $0.04 overages
- ✓Strong AWS + IAM Identity Center enterprise path
- ✓VS Code-compatible editor reduces migration cost
- ✓Multi-surface subscription (IDE, CLI, web) on one credit pool
- ✓Free tier remains usable for light evaluation (50 credits + open weights)
Cons
- ✗Credit forecasting is harder than flat unlimited-feeling plans, multi-agent runs can burn Pro quickly
- ✗Spec overhead slows pure vibe-coding compared with Cursor-style flows
- ✗Model lineup and credit multipliers shift with vendor announcements
- ✗Region/model availability caveats on premium Claude models
- ✗Team features and identity are AWS-shaped, less ideal if you want zero Amazon footprint
Community & Ecosystem
Kiro discussion spans AWS blogs, RedMonk analysis (including 2026 iOS/remote-session framing), indie reviews, and X threads about credit burn when running several agents at once. Student promo posts circulate alongside Pro Max/Power upgrade stories. Ecosystem connectors include CLI automation and ACP-compatible editors. Compared with Cursor's social density, Kiro's loudest advocates are teams that care about specs, governance, and AWS delivery, not only autocomplete speed.
Frequently Asked Questions about Kiro
About the Author
Expert Tips for Kiro
Practitioners on X report that multi-agent Kiro CLI sessions can burn through hundreds of dollars of credits in days, treat Power/$0.04 overages as real OpEx, not edge cases.
Official pricing docs are unusually explicit about credit multipliers by model; teams that standardize on Auto for routine work and premium Claude only for hard tasks keep Pro/Pro+ viable.
Related Technologies & Concepts
Related Topics
Sources & References
Fact-CheckedLast verified: Aug 9, 2026
Key Verified Facts
- Kiro individual pricing is Free $0 (50 credits), Pro $20 (1,000), Pro+ $40 (2,000), Pro Max $100 (5,000), Power $200 (10,000), with $0.04 add-on credits.[cite-2]
- Free tier includes open-weight models such as Qwen3 Coder Next, DeepSeek v3.2, and MiniMax 2.1 plus Claude Sonnet 4.5 with rate limits; paid plans list Auto, Claude Sonnet 4.6, and Claude Opus 4.8 among premium models.[cite-2]
- Team plans add consolidated billing, usage analytics, SAML/SCIM via AWS IAM Identity Center, and organizational controls at the same per-user price points as individual tiers.[cite-2]
- Independent 2026 comparisons place Kiro Pro at $20/mo with a $100 Pro Max tier against Cursor's Pro/Ultra ladder.[cite-4, cite-6]
- 1
- 2Kiro pricing
Accessed Aug 9, 2026
"Free 50 credits; Pro $20; Pro+ $40; Pro Max $100; Power $200; $0.04 add-ons; model footnotes."
- 3
- 4Kiro vs Cursor 2026 comparison
Accessed Aug 9, 2026
"Independent table of Free/Pro/Pro+/Pro Max/Power vs Cursor tiers."
- 5RedMonk: Kiro and the AI IDE market
Accessed Aug 9, 2026
"June 2026 analysis including iOS/remote development framing."
- 6Kiro review 2026 (pricing verification)
Accessed Aug 9, 2026
"Documents Free/Pro/Pro+/Pro Max/Power and model notes."
- 7X: Kiro credit burn discussion
Accessed Aug 9, 2026
"Practitioner notes multi-agent credit spend on Kiro CLI."
- 8X: Kiro student Pro promo chatter
Accessed Aug 9, 2026
"Community post about education offers for Kiro Pro."
Research queries: Kiro AWS IDE pricing 2026; Kiro Pro Max Claude models