Back to vendors
K

Kiro

Also known as: Kiro CLI, Kiro Crew, Kiro Web

Visit site
Entry priceFree (50 credits) · Pro $20/user/moFull pricing detail

AWS's spec driven agentic development environment across IDE, CLI, Web and Mobile, where structured requirements and design documents are approved before code is written, with Powers plugins, agent hooks and deep enterprise governance.

Kiro is AWS's spec driven agentic development environment, built on Code OSS and generally available in 2026 as the successor to Amazon Q Developer, which reaches end of support in April 2027. Its central claim is engineering rigour: rather than going from prompt to code, Kiro first generates structured requirements, design and task documents that the developer reviews and approves, after which the agent implements against that plan. The spec becomes the source of truth in the repository and the code becomes a build artifact.

One unified agent harness fronts four surfaces, the IDE, a terminal CLI, a Web preview and Mobile, with configuration shared through a .kiro directory so a spec started in the IDE can continue in the CLI and hand off to the Web agent. Steering files carry project standards and conventions across every surface with always, auto, fileMatch and manual activation modes, and AGENTS.md is supported at global and workspace scope. Agent hooks fire on file save, commit and repository events to run tests, update documentation or validate work against the requirements document.

Extensibility runs through MCP, with local and remote servers, hot reloading configuration and per agent definitions, and through Powers: one click installable plugins that bundle MCP tools, skills and knowledge, built on the open Agent Plugins specification whose maintainers include Amazon, Cursor, Microsoft, OpenAI and Vercel. Custom agents are declared in JSON with capability based permission rules that allow or deny specific shell commands and file paths.

Enterprise governance is unusually deep for a coding agent: identity provider connection through IAM Identity Center, organization permission policies, model governance, API key and web tool governance, an MCP allow list registry that fails closed if the governance API is unreachable, usage monitoring, managed updates, and AWS PrivateLink VPC endpoints. Models are served through Amazon Bedrock with customer selection, a configurable reasoning effort level and Auto routing. Pricing is credit based with a free tier, Pro at twenty dollars, Pro+ at forty and Power at two hundred.

Vendor details

Canonical URL

https://kiro.dev

Category

Coding agent

Subcategory

Coding — agentic IDE (spec-driven)

Company status

first party product

Use cases & customers

Primary use cases

spec-driven developmentagentic codingsoftware development

Target customers

developers

Deployment options

IDE (desktop, Code OSS based)CLI (including headless and ACP)Web (preview, sandboxed sessions)Mobile (preview)Crew (continuous 24/7 operation)AWS PrivateLink VPC endpoints (enterprise)

Integrations

MCP client support across local stdio and remote streamable HTTP servers with hot reloading config, a server directory, tool search and per agent server definitions, plus an enterprise MCP registry allow list. Powers bundle MCP tools, skills and knowledge into one click installable plugins following the open Agent Plugins specification, browsable at kiro.dev/powers. Deep native AWS integration reaches IAM, S3, Lambda and Bedrock without manual configuration, and agent hooks fire on file save, commit and repository events.

In practice

Agents that jump straight to code drift from what you actually asked for. Kiro is spec-driven, working from a written spec so the implementation matches the intent.

You want the same agentic IDE on your laptop and on the go. Kiro runs across CLI, web, and iOS, so the work follows you off the desktop.

You want strong models without managing providers. Kiro offers Claude through Bedrock alongside OpenAI, DeepSeek and MiniMax models, and Auto routing picks the right model for each task.

Agentic Index coverage score

13.0 / 14 capabilities · 93%

Integrations & Tool Calling Full

MCP client support spans local stdio and remote streamable HTTP servers with hot reloading configuration, a server directory, tool search and per agent server definitions, plus built in tools and deep AWS integration reaching IAM, S3, Lambda and Bedrock natively; Powers bundle MCP tools with skills for one click installation.

Sourcekiro.dev/docs/mcp, docs/tools and docs/powersread 2026-08-30

Workflow Orchestration Full

Spec driven development chains a multi step workflow: a prompt produces structured requirements, design and task documents which the developer reviews and approves, after which the agent implements across files against the task list; hooks chain further agentic work on repository events, custom agents can be defined per project, and Crew coordinates continuous multi agent operation.

Sourcekiro.dev/docs/specs, docs/hooks, docs/custom-agents and docs/crewread 2026-08-30

Knowledge Grounding & RAG Full

Crew Knowledge is a curated library of documents, folders and URLs the customer feeds the agent, ingested and searchable, and the CLI's knowledge management stores and searches context across chat sessions in knowledge bases. Tree-sitter and LSP code intelligence, steering files and specs add per task context.

Sourcekiro.dev/llms.txt (Crew Knowledge, CLI knowledge management) and docs/tools/code-intelligence; kiro.dev/docs/crew/features/knowledgeread 2026-09-29

Human Oversight & Guardrails Full

Capability based permission rules allow or deny specific shell commands and file paths per agent, with organization level permission policies set in the enterprise console, trusted command configuration, .kiroignore file exclusions, and checkpoints allowing rewind of agent work. The spec workflow itself holds implementation until a person approves the requirements and design document.

Sourcekiro.dev/docs/permissions, docs/enterprise/governance/permissions and docs/specsread 2026-08-30

Security, Identity & Governance Full

Enterprise governance covers identity provider connection with IAM Identity Center, permission policies, per organization MCP enable or disable with an allow list registry that fails closed if the governance API is unreachable, model governance, API key governance, web tool governance, monitoring and tracking, and managed updates. Privacy and security controls cover data protection, encryption, compliance validation, infrastructure security, IAM permissions, data perimeters and VPC endpoints via AWS PrivateLink.

Sourcekiro.dev/docs/enterprise/governance and docs/privacy-and-securityread 2026-08-30

Observability & Auditability Partial

Enterprise administrators get a monitor and track view of organization usage and activity, specs leave requirements, design and task documents as a durable record of intent, and checkpoints allow rewinding to an earlier point in a session. Kiro does not describe a per agent execution trace or an audit log that reconstructs tool calls and reasoning.

Sourcekiro.dev/docs/enterprise/monitor-and-track and docs/checkpointsread 2026-08-30

Memory & State Persistence Full

Kiro Web keeps a memory of what the agent learns from the task creator's feedback, shown in Settings where each memory can be deleted. Kiro Crew keeps six memory layers across sessions (preferences, projects, recent history with decay, semantic key value memory, episodic memory with vector search and time decay, and learned lessons). Specs and steering files hold standing instructions.

Sourcekiro.dev/docs/web/memory and docs/crew/features/memoryread 2026-09-29

Deployment & Data Residency Full

Enterprise deployment covers supported AWS regions, VPC endpoints via AWS PrivateLink so traffic need not traverse the public internet, firewall, proxy and data perimeter configuration, deployment options and IAM integration. The client itself runs locally on desktop, CLI, web and mobile surfaces.

Sourcekiro.dev/docs/privacy-and-security/vpc-endpoints, docs/enterprise/supported-regions and docs/enterprise/deployment-optionsread 2026-08-30

Prebuilt Agents, Templates & Packs Full

Powers are installable plugins bundling MCP tools, skills and knowledge into a single package, browsable in the client or at kiro.dev/powers with one click install and dynamic context activation, built on the Agent Plugins specification with maintainers from Amazon, Cursor, Microsoft, OpenAI and Vercel; Agent Skills ship as standalone portable instruction packages, and a Power Builder converts older formats.

Sourcekiro.dev/docs/powers and docs/skillsread 2026-08-30

Triggers & Channel Coverage Full

One agent harness fronts four surfaces, IDE, CLI, Web and Mobile, with configuration shared through .kiro so a spec started in the IDE continues in the CLI and hands off to the Web agent. Agent hooks fire on file save, commit and repository events, the CLI adds headless mode, ACP and voice, Crew runs continuously around the clock, and the Web surface adds autonomous mode and automations.

Sourcekiro.dev/docs, docs/hooks and docs/cli/headlessread 2026-08-30

Model Flexibility & Routing Full

Users choose among models from Anthropic (Claude Opus, Sonnet, Haiku and Fable), OpenAI (GPT-5.6 Sol, Terra and Luna), DeepSeek and MiniMax with a configurable reasoning effort, or let Auto route by task, and enterprise administrators govern which models their organization may use.

Sourcekiro.dev/docs/models/available-models and docs/enterprise/governance/modelread 2026-09-29

APIs, SDKs & MCP Extensibility Full

MCP servers are configurable per workspace, globally or per agent with hot reload, custom agents are defined declaratively in JSON with tools, permissions, resources and lifecycle hooks, Powers follow the open Agent Plugins specification and can be authored and published, and the CLI exposes headless mode and ACP for programmatic and third party client use. Kiro also publishes machine readable schemas and a markdown view of every documentation page.

Sourcekiro.dev/docs/mcp/configuration, docs/custom-agents/configuration-reference and docs/cli/acpread 2026-08-30

Testing, Debugging & Optimization Partial

Hooks run tests, validate task completion against the spec's requirements document and trigger security scans on file save, commit or repository events, and the spec workflow supports verifying generated code against the written requirements. Those checks test the customer's code against intent, and there is no harness, replay or regression suite for the agent's own behavior.

Sourcekiro.dev/docs/hooks and docs/specsread 2026-08-30

Browser & Computer Use Full

Kiro Crew operates a real browser for the user, navigating, clicking, typing, submitting forms, inspecting accessibility and taking screenshots, through a native Browser tool in the packaged desktop app with a live panel the user can take over, or by attaching to the user's own Chrome when chosen, with approvals applying.

Sourcekiro.dev/docs/crew/features/browserread 2026-09-29

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Recent platform changes

2026-09-30·Workflow orchestrationVerified

Kiro shipped Workflows, an opt in feature across its web, IDE and CLI. Kiro proposes a reusable multi step plan from a described outcome, runs each step in its own background agent session with loops and parallel branches, and pauses when a step needs input. Users can steer, pause, retry or stop steps and save the process as a recipe.

Bears on: Workflow orchestration

View source
2026-09-24·Workflow orchestrationVerified

Kiro Crew 0.7.0 added a durable queue for accepted subagent work, Task Runner steps, and workflow calls, restoring queued work after Gateway restarts. The release also added AWS Fargate to its optional remote crew preview, reaching tasks through AWS Systems Manager without an inbound network rule. Fargate tasks have a default six hour lifetime and a fixed limit of ten concurrent tasks.

Bears on: Deployment / data residency

View source
2026-09-14·Agent capabilityVerified

Kiro raised the context window for GPT-5.6 Sol, Terra and Luna to one million tokens, up from 272,000.

Bears on: Agent capability

View source
View all 10 changes for Kiro →Tracked since Jun 2026 · Verified from public vendor sources

Pricing

Free (50 credits) · Pro $20/user/mo

credits

Free tierTrial available

Included quota

Pro (entry paid, $20/mo): 1,000 credits/mo plus premium models (Auto, Claude Sonnet 4.6, Claude Opus 4.8) and the full Kiro feature set; Free is 50 credits/mo on open-weight + Sonnet, rate-limited.

What is public

Kiro (AWS's agentic IDE) publishes five per-user tiers: Free ($0, 50 credits/mo), Pro $20 (1,000 credits), Pro+ $40 (2,000), Pro Max $100 (5,000, added June 2026), and Power $200 (10,000). Overage is $0.04 per extra credit (opt-in, off by default). New users get 500 bonus trial credits (expire in ~14 days) and a $20 first-upgrade bonus.

Billing mechanics

Credit-metered: a credit is a unit of work, consumed fractionally (0.01 increments) by prompt complexity — simple edits <1 credit, spec tasks >1. Each paid tier bundles a monthly credit block; credits don't roll over and are per individual user (no shared team pool yet). Model choice affects burn (Sonnet 4 costs ~1.3x the 'Auto' agent for the same task).

Cost watchouts

$0.04/credit overage on heavy spec/agent workflows, non-rollover credits, per-individual subscriptions (no pooled team billing yet), GovCloud ~20% higher, premium-model burn rate

Variable cost rationale

Credit metering plus opt-in $0.04 overage means heavy daily/agentic use can push real cost well past the flat tier fee

Additional watchouts

The flat tier is just the floor — overage and non-rollover credits drive real cost; no pooled team billing yet (each dev needs a subscription)

Overage / add-ons

Above the monthly credit block, usage bills at $0.04 per additional credit — off by default and opt-in in Settings; once enabled it stays on while you remain on a paid plan; billed at month-end.

Sales call required

No, self serve available

Free / trial

Free tier

Lowest paid plan

Pro $20/user/month

Commercial notes

$20 first-upgrade bonus; AWS Startups program offers up to 1 yr free Kiro Pro+ for early-stage/VC-backed startups; AWS Activate Credits usable on Kiro; works across IDE/CLI/web/CI-CD; migrated from Amazon Q Developer

Key ambiguities

Exact credit cost per prompt/spec task varies by complexity and model and isn't fixed; the trial-credit window is stated as ~14 days but some pages reference 30

Cancellation / refund

Monthly subscriptions; upgrades effective immediately (prorated), downgrades take effect next month; a mid-month downgrade still charges the higher tier for that month; per-user billing

Support SLA / resale

Enterprise tier adds centralized billing, SSO (SAML/SCIM via AWS IAM Identity Center), usage analytics, and org management; standard support otherwise

Missing data

Precise per-task credit costs (complexity/model-dependent) aren't enumerated; team/pooled billing is 'coming soon' and unpriced

Agentic Index verified 2026-09-29

Alternatives to Kiro

The closest documented capability profiles to Kiro among coding agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.

  • Kilo Code13.0 / 14Matches Kiro across all 14 documented capabilities
  • GitHub Copilot13.5 / 14Fuller documented coverage on Observability & AuditabilityKiro vs GitHub Copilot →
  • Refact.ai12.5 / 14A lighter documented profile than Kiro
  • Cognition13.0 / 14Fuller documented coverage on Observability & Auditability
  • OpenAI Codex13.0 / 14Fuller documented coverage on Observability & Auditability
  • Warp13.0 / 14Fuller documented coverage on Observability & Auditability

Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded

Contact us

Found a vendor we missed? Have feedback on the index? We'd love to hear from you.