Pricing guide
Kiro Pricing (2026): Kiro Cost Per Month, Free Tier, Credits and Enterprise Plans
Every Kiro tier and what it costs, plus the part most pricing articles leave out: a credit is a variable unit, and the model you pick multiplies it by anywhere from 0.05x to 2.4x. Figures read from Kiro's own pricing page, FAQ and model docs in August 2026.
No card to start · Review-first: the agent never merges
Pick a task
Plan
- planning
Files changed
Test run
Pull request
You review and merge. Agentcode never merges on its own.
The short answer
Kiro costs $0 for the free tier with 50 credits a month, then $20 per user for Pro with 1,000 credits, $40 for Pro+ with 2,000, $100 for Pro Max with 5,000 and $200 for Power with 10,000. Extra credits cost $0.04 each. The number that actually decides your bill is not the tier, it is the model multiplier: the same task costs 0.05x on Qwen3 Coder Next and 2.4x on GPT-5.6 Sol, so one 1,000-credit plan can be worth 48 times more work depending on what you pick. Every paid tier includes credits at exactly $0.02 each, which makes overage precisely double the included rate. If you would rather have a flat bill and an agent that opens a reviewable pull request on the GitHub or GitLab repo you already have, Agentcode is $29 a month with no meter.
Kiro pricing plans and what each costs
Kiro publishes five tiers. They scale in a perfectly straight line: every paid plan includes credits at $0.02 each, so moving up buys headroom rather than a better rate. Team plans use the same per-user prices and add admin controls.
Swipe to see more
| Plan | Price | Included credits | Best for |
|---|---|---|---|
| Free | $0 per month | 50 credits | Trying Kiro and deciding if spec-driven work suits you |
| Pro | $20 per user, per month | 1,000 credits | One developer using an agent daily on a normal codebase |
| Pro+ | $40 per user, per month | 2,000 credits | Heavy daily use, or anyone who keeps hitting the Pro ceiling |
| Pro Max | $100 per user, per month | 5,000 credits | Full-time agentic work, large refactors, long spec sessions |
| Power | $200 per user, per month | 10,000 credits | The heaviest individual users and credit-hungry teams |
Prices and credit allowances as published by Kiro and checked in August 2026. Add-on credits cost $0.04 each. Confirm on kiro.dev/pricing before you buy.
What does a Kiro credit actually buy?
This is the table that decides your bill, and it is the one missing from most Kiro pricing write-ups. Kiro charges credits for a unit of work, then multiplies by the model you selected. Auto is the 1.0x baseline. Effective cost below assumes the $0.02 per credit you pay on any paid tier.
Swipe to see more
| Model | Credit multiplier | Effective cost | What that means |
|---|---|---|---|
| Qwen3 Coder Next | 0.05x | $0.001 | Cheapest option by a wide margin. 20,000 credits of work inside a $20 Pro plan. |
| GPT-5.6 Luna | 0.1x | $0.002 | The light GPT tier, ten times cheaper than the Auto baseline. |
| MiniMax M2.1 | 0.15x | $0.003 | Open-weight model, useful for bulk mechanical edits. |
| DeepSeek 3.2 | 0.25x | $0.005 | A quarter of baseline cost, so Pro stretches to 4,000 credits of work. |
| MiniMax M2.5 | 0.25x | $0.005 | Same rate as DeepSeek 3.2. |
| Claude Haiku 4.5 | 0.4x | $0.008 | The cheap Claude. Reasonable for small, well-specified edits. |
| GLM-5 | 0.5x | $0.010 | Half baseline. |
| Auto (default) | 1.0x | $0.020 | Kiro routes across frontier and specialized models for you. This is the baseline every other multiplier is measured against, and the cheapest way to get frontier-class output. |
| GPT-5.6 Terra | 1.0x | $0.020 | Same rate as Auto. |
| Claude Sonnet 4.0 to 5 | 1.3x | $0.026 | Every Sonnet version bills at 1.3x, so forcing Sonnet costs 30 percent more than letting Auto choose. |
| Claude Opus 4.5 to 5 | 2.2x | $0.044 | Every Opus version bills at 2.2x. A 1,000-credit Pro plan becomes roughly 455 credits of Auto-equivalent work. |
| GPT-5.6 Sol | 2.4x | $0.048 | The most expensive model on the platform, 48 times the rate of Qwen3 Coder Next. |
Read the top and bottom rows together and the point lands: a $20 Pro plan is worth 20,000 credits of work on Qwen3 Coder Next and about 417 on GPT-5.6 Sol. That 48-fold spread is larger than the gap between the cheapest and most expensive plan Kiro sells, which is why "what does Kiro cost" has no single answer. Multipliers read from kiro.dev/docs/models in August 2026; effective costs are our arithmetic at the $0.02 included rate.
How Kiro credits work, and where the bill surprises people
Kiro moved to a single unified credit pool, so spec work, vibe prompts and agent hooks all draw from the same balance. Four mechanics do most of the damage to a budget.
A credit is a unit of work, not a request
Kiro defines a credit as a unit of work performed in response to your prompt, and meters it to two decimal places. The smallest charge possible is 0.01 credits. A one-line edit can cost a fraction of a credit; executing a full spec task usually costs more than one. That is fairer than charging a flat rate per message, but it means you cannot convert your plan into a number of prompts in advance. Two developers on identical $20 plans can get very different amounts of work done.
The model multiplier is the biggest lever on your bill
Picking the model is the single largest cost decision in Kiro, and it is buried in a dropdown. The same task costs 0.05x on Qwen3 Coder Next and 2.4x on GPT-5.6 Sol, a 48-fold spread. Auto sits at the 1.0x baseline and routes across frontier models for you, which is why Kiro made it the default: forcing Sonnet costs 30 percent more than Auto for the same job, and forcing Opus costs 120 percent more.
Overage costs exactly double the included rate
Every paid tier works out to $0.02 per included credit: $20 for 1,000, $40 for 2,000, $100 for 5,000, $200 for 10,000. Extra credits cost $0.04 whether you buy an add-on pack or let a team overage bill at month-end. So the credits you run past your allowance are precisely twice the price of the ones you planned for, which is the strongest argument for sizing your tier up rather than topping up every month.
Monthly credits expire, purchased credits do not
Your plan credits reset at the start of each billing month and unused ones do not roll over. Add-on credits behave differently: they carry forward and expire twelve months after purchase. If your work is lumpy, a smaller plan plus occasional add-on packs wastes less than a large plan you only use in busy months, even though the add-on rate is double.
None of this makes Kiro expensive. It makes Kiro variable, which is a different problem: you can forecast a seat, and you cannot forecast a meter without watching it. We keep a running flat vs credit-metered pricing comparison across every major tool, because this billing shape is now the norm rather than the exception.
Which Kiro plan do you actually need?
Free
$0 · 50 creditsFifty credits a month, which is genuinely a trial rather than a working allowance: a single spec task can consume more than one credit, so a free month is a few dozen real pieces of work at most. Every model is available, including the Claude and GPT families, with limits. If you upgrade through a social login or an AWS Builder ID, Kiro credits $20 toward the first bill. The free tier is not offered in AWS GovCloud.
Pro
$20 · 1,000 creditsThe tier most individuals land on. A thousand credits at the Auto baseline covers steady daily use for a single developer, though what that means in tasks depends entirely on the model you select and how complex your prompts are. Run out and you buy add-on packs at $0.04 a credit, starting at $5 for 125 credits. There is no annual discount; Kiro bills on the first of each calendar month.
Pro+
$40 · 2,000 creditsExactly double Pro for exactly double the credits, so there is no volume discount for stepping up. The reason to move is not a better rate, it is avoiding the interruption of running dry mid-task. If you bought add-on credits on Pro two months running, Pro+ costs the same and saves you the friction.
Pro Max
$100 · 5,000 creditsFive thousand credits suits someone whose main workflow is the agent rather than the editor, especially on premium models where the multiplier eats the allowance fast. At the 2.2x Opus multiplier this is worth roughly 2,270 credits of Auto-equivalent work, which is a useful way to sanity-check whether you need it.
Power
$200 · 10,000 creditsThe ceiling tier. Same $0.02 per included credit as every other paid plan, so you are buying headroom rather than a better price. For most teams the honest comparison is not Power against Pro Max, it is whether a metered plan of any size is the right shape for your spend at all.
Is Kiro replacing Amazon Q Developer?
Yes, and if that is why you are reading a Kiro pricing page, the budget change matters more than the feature change. AWS blocked new Amazon Q Developer signups on May 15, 2026, and the IDE plugins plus paid Pro subscriptions reach end of support on April 30, 2027, with the Java and .NET code transformation features retiring alongside them. AWS points developers at Kiro, and for the transformation work at AWS Transform.
Amazon Q Developer Pro was a mostly-flat $19 per user per month. Kiro Pro is $20 per user per month plus a meter. On paper that is a dollar. In practice you are moving from a seat you could put in a spreadsheet to a balance that moves with model choice and prompt complexity, so budget a busy month rather than an average one. Our Amazon Q Developer pricing page carries the full end-of-support timeline, and the Amazon Q Developer vs Kiro breakdown walks through what the migration actually involves.
Worth stating plainly, because the retirement news causes confusion: Kiro itself has no announced sunset. It is the product AWS is investing in, not another one being wound down.
Kiro vs a flat-priced pull request agent
Kiro is a full agentic IDE organized around approving a spec before code gets written; Agentcode is an async agent that opens a reviewable pull request on a repository you already have. Different jobs, so here is where each is genuinely ahead.
Swipe to see more
| Kiro | Agentcode | |
|---|---|---|
| Billing shape | Credit-metered. Five tiers from $0 to $200 per user, monthly only, overage at $0.04 per credit | Flat $29 a month, no usage meter and no overage line |
| Cost predictability | Varies with model choice and prompt complexity; a busy month costs more | A busy month and a quiet month cost the same |
| Where it runs | A standalone IDE built on Code OSS, plus a CLI and a web interface with cloud sandboxes | On your repository. Nothing new to install and no editor to switch to |
| How work starts | Spec-driven: prompt becomes requirements, then design, then sequenced tasks you approve | You describe the task, the agent plans it and opens a pull request |
| Git hosts | GitHub and GitLab both supported | GitHub and GitLab both supported |
| Model choice | Eighteen models, each with its own credit multiplier you have to manage | Model selection handled for you, with no cost consequence to weigh |
| Review model | You approve the spec and tasks inside the IDE as the agent works | Review-first: the agent opens a pull request and never merges on its own |
| Free option | Yes, 50 credits a month, not available in GovCloud | Start without a card |
| Best for | Teams who want structured, spec-first development and are happy managing a meter | Teams who want a reviewable change on an existing repo and a bill they can forecast |
On the merits, Kiro wins when you want structure. Spec-driven development, where the agent turns a prompt into requirements and a design and a sequenced task list you sign off before a line is written, is a genuinely better fit for large, ambiguous work than firing prompts at an editor. Eighteen models and a proper CLI and cloud sandboxes are real advantages, and unlike some rivals it supports GitLab as well as GitHub, so we will not pretend otherwise. Where a flat plan wins is forecasting and friction: Agentcode runs on the repo you already have for one monthly price, with no new IDE to adopt, no multiplier to think about, and no month where the bill doubles because someone left Opus selected.
Kiro pricing: your questions answered
How much does Kiro cost?
Kiro costs $0 on the Free tier for 50 credits a month, $20 per user for Pro with 1,000 credits, $40 for Pro+ with 2,000, $100 for Pro Max with 5,000, and $200 for Power with 10,000. Extra credits cost $0.04 each. Every paid tier works out to the same $0.02 per included credit, so higher plans buy headroom rather than a discount.
Is Kiro free?
Yes, Kiro has a permanent free tier with 50 credits a month and access to every model, including the Claude and GPT families, with limits. Fifty credits is a trial rather than a working allowance, because a single spec task can consume more than one credit. The free tier is not available in AWS GovCloud, and upgrading through a social login or AWS Builder ID adds a $20 credit.
How do Kiro credits work?
A Kiro credit is a unit of work done in response to your prompt, metered to two decimal places with a 0.01 credit minimum. Simple edits cost under a credit, and executing a spec task usually costs more than one. Your total is then multiplied by the model you chose, from 0.05x on Qwen3 Coder Next to 2.4x on GPT-5.6 Sol, with Auto as the 1.0x baseline.
What happens when you run out of Kiro credits?
On an individual plan, work stops until you buy an add-on pack, priced at $0.04 per credit and starting at $5 for 125 credits. On Team and Enterprise plans an admin can enable overages so people keep working uninterrupted, with the extra usage billed at $0.04 per credit at month-end. Add-on credits roll over and expire twelve months after purchase.
Do Kiro credits roll over?
Your monthly plan credits do not roll over. They reset at the start of each billing month and anything unused is lost. Add-on credits you purchase separately are the exception: those carry forward and expire twelve months from the date of purchase. If your workload is uneven, that difference is worth planning around before you pick a tier.
What is Kiro enterprise pricing?
Kiro Team and Enterprise plans use the same per-user prices as the individual tiers, $20, $40, $100 and $200, and add consolidated billing, usage analytics, an organizational dashboard, SAML and SCIM single sign-on, and enterprise security controls. Admins can enable pay-per-use overage at $0.04 per credit. AWS GovCloud pricing runs roughly 20 percent higher and has no free tier.
Does Kiro have annual billing?
No. Kiro bills monthly and processes paid subscriptions on the first day of each calendar month, with no published annual discount. That is unusual in this category, where most vendors shave 15 to 20 percent for an annual commitment. It does mean you can move between tiers or leave without being locked into a year.
Is Kiro replacing Amazon Q Developer?
Yes, for the coding assistant. AWS blocked new Amazon Q Developer signups on May 15, 2026 and its IDE plugins and paid Pro subscriptions reach end of support on April 30, 2027, with AWS pointing developers to Kiro. The billing shape changes with the move: Q Developer Pro was a mostly-flat $19 seat, and Kiro is metered.
How does the Kiro $20 sign-up credit work?
It is prorated, and the date you upgrade decides what it is worth. Kiro processes subscriptions on the 1st of each calendar month, so upgrading on the 1st credits the full $20 and Kiro's own example puts a June 15 upgrade at $20 times 15/30, which is $10. It requires a social login or AWS Builder ID, not AWS Identity Center.
Can an enterprise get the Kiro sign-up bonus?
Usually not. Kiro restricts the $20 first-upgrade credit to accounts using a social login or an AWS Builder ID and rules out AWS Identity Center explicitly, which is how most enterprises federate their developers. If you are running a paid pilot through Identity Center, budget without the bonus and without the free tier if you are in GovCloud.
Is Kiro worth $20 a month?
It is worth it if spec-driven development matches how you work and you are willing to manage the model multiplier, since Auto at 1.0x makes 1,000 credits go considerably further than Opus at 2.2x. It is a weaker fit if you need a predictable bill, do not want to adopt another IDE, or mainly want reviewable changes landing on a repository you already have.
Last updated: August 2026
Weighing Kiro against the field?
These pages compare pricing tool by tool: Amazon Q Developer pricing, Cursor pricing, GitHub Copilot pricing, Claude Code pricing, Devin pricing, Replit pricing and Augment Code pricing.
Or read the head-to-head on Kiro vs Claude Code, where the credit ceiling meets a meter with no ceiling at all, and on Kiro vs Cursor, step back to AI coding tool pricing across every tool, the best AI for coding in 2026, or what an AI coding agent does before you weigh the bill.
A flat bill, on your repo
Connect a GitHub or GitLab repository you already have, describe a task, and read the pull request it opens. Same price whether you run it once or a hundred times, with no credits to top up and no multiplier to manage.