If you opened your IDE this month and tried to spin up a fresh Amazon Q Developer login, you already know: that door is closed. As of May 15, 2026, AWS stopped accepting new Q Developer signups — no new Free Tier via Builder ID, no new Pro subscriptions through the Console. The plugin still works if you’re already on it, but the writing is on the wall. Amazon Q Developer’s IDE story is ending, and AWS wants you on Kiro instead.
This isn’t a rumor or a “deprecated, use at your own risk” footnote. AWS published a hard timeline. If your team built workflows around Q Developer’s inline completions, chat, or agents, you’ve got a real migration on your hands — and a year to do it well or do it in a panic.
I’ve spent the last couple of weeks digging into what’s actually going away, what survives, and which replacement makes sense depending on how you work. Here’s the honest version.
What’s ending, and the dates that matter
The announcement (the AWS DevOps blog post from late April 2026) lays out three checkpoints. Write these down:
- May 15, 2026 — New signups blocked. Free Tier account creation through Builder ID in the IDE plugins, and new Q Developer subscriptions via the AWS Console, are both shut off. This already happened.
- May 29, 2026 — Opus 4.6 gets pulled from Q Developer Pro. If your prompts were tuned around that model, you’ll feel the change first here, weeks before anything else breaks.
- April 30, 2027 — Full end of support for the Q Developer IDE plugins (VS Code, JetBrains, Eclipse, Visual Studio) and paid subscriptions. This is the real cliff. Twelve months from the signup freeze.
A few things that surprised me in a good way. If you have an active Q Developer Pro subscription, you can still add new users to it — the May 15 freeze only blocks brand-new accounts, not seats on an existing plan. So a team that’s already paying isn’t locked out of onboarding the new hire next quarter. And the plugins stay published in all four IDE marketplaces through the transition, with a deprecation banner pointing at Kiro. AWS says critical bugfixes keep flowing to existing users until the cutoff. So nothing turns into a pumpkin overnight.
What actually survives
Here’s the part people miss in the panic headlines: Amazon Q in the AWS Management Console isn’t going anywhere. Neither are the other first-party AWS surfaces — the documentation site, the Console mobile app, and Q Developer in chat apps (Slack and Microsoft Teams). Those keep running, and your current subscription benefits there stay intact.
So “Amazon Q Developer is dead” is wrong. What’s dying is the IDE coding assistant — the inline completions and in-editor agent you used while writing code. The “ask Q about your AWS bill in the Console” experience is fine. Keep that mental separation, because it changes what you actually need to replace. You’re not replacing all of Q. You’re replacing the part that lived in your editor.
Should you just follow AWS to Kiro?
AWS’s answer to “what now?” is Kiro, its agentic IDE built around spec-driven development. The pitch: instead of going prompt-to-code, Kiro writes requirements documents, a design doc, and a structured task list first. You review and approve the plan, then it writes code. It layers on hooks (automated actions on events), steering files (project-level context), subagents, and reusable “Powers.”
I’ll be straight about my take. The spec-first model is genuinely good for greenfield features and for teams that hate the “the AI wrote 400 lines and I don’t know what half of it does” problem. It forces a pause. For a refactor on a sprawling legacy codebase, or for the quick “rename this and fix the call sites” tasks that make up most of a real workday, the ceremony can feel like overhead. It’s a different philosophy from the autocomplete-on-steroids tools, not strictly an upgrade.
Pricing as of May 2026: Kiro Free gives you 50 interactions a month (Claude Sonnet 4.5 plus open-weight models like Qwen3 Coder Next, DeepSeek v3.2, MiniMax 2.1 if you signed in with a social login or Builder ID). Paid starts at Kiro Pro at $19/month, with Pro+ and Power tiers above that. One thing worth knowing if you’re a heavy user: newer frontier models — Opus 4.7, for instance — launch on Kiro, not on the sunsetting Q plugin. And if you’re in AWS GovCloud (US), expect roughly 20% higher pricing and no Free tier.
The honest case for Kiro: you’re AWS-heavy, you want the vendor-blessed path, and the spec workflow clicks with your team. The honest case against: you just want fast completions in your existing editor and don’t want to relearn a tool, or 50 free interactions a month evaporates by Tuesday and you’re not sure the model lineup justifies switching from whatever else you could pick.
The alternatives worth a serious look
Following AWS isn’t the only move. A forced migration is actually a decent moment to reconsider the whole category, since you’re already paying the switching cost. Here’s how the main contenders stack up.
GitHub Copilot — the safe default
Copilot is where most teams land, and for good reason: it’s mature, it’s everywhere, and the IDE support is the best in the category. The catch is that pricing is in flux. Plans today run Free ($0, ~2,000 inline suggestions/month), Pro at $10/month, Pro+ at $39/month, and Business at $19/user/month.
But starting June 1, 2026, Copilot moves to usage-based billing. Instead of counting “premium requests,” every plan ships with a monthly allotment of GitHub AI Credits, and paid plans can buy more. Usage is metered on token consumption — input, output, and cached tokens — at each model’s API rate. Monthly Pro and Pro+ users migrate automatically. If your team runs a lot of agent calls against expensive models, model your costs before assuming $10 covers it. The headline price didn’t change; how far it stretches did.
For most teams replacing Q Developer, Copilot is the lowest-risk pick. Broad model choice, deep editor integration, and you’re not betting on a niche vendor surviving.
Gemini Code Assist — great free tier, but read the fine print
Gemini Code Assist has the most generous free tier I’ve seen: 6,000 code-related requests per day and 240 chat requests per day for individuals, plus Google’s large context window, which is handy on big files. Standard runs around $19–23/user/month and Enterprise around $45–54, depending on the source and billing terms.
Here’s the caveat that should give you pause before you commit your whole team: Google is unifying its tools into a platform called Antigravity, and the Gemini Code Assist IDE extensions and Gemini CLI stop serving requests for the individual, Google AI Pro, and Google AI Ultra tiers starting June 18, 2026. So you’d be migrating off one sunsetting tool onto another tool that’s mid-transition. That’s not automatically a dealbreaker — Antigravity is the destination, not a dead end — but if your whole reason for leaving Q was “I’m tired of tools getting yanked,” go in with eyes open. Verify the current state on Google’s docs before you bet a team on it.
Continue.dev — when you want to own the stack
If data privacy or vendor lock-in is what burned you here, Continue.dev is the most interesting option. It’s open-source, the core extension is free, and it was built local-first around Ollama and LM Studio. You can wire it to a locally hosted model and run a genuinely air-gapped setup where no code leaves your network. Their hosted tiers go Solo ($0), Team ($10/developer/month), and Enterprise (custom, with governance and self-hosting).
The trade-off is real, though. Local models that run on a developer laptop aren’t Claude or GPT-class. You’re trading raw capability for control and zero per-token cost. For a lot of completion and boilerplate work that’s a fine trade; for hard reasoning over a large codebase, you’ll notice the gap. This is the pick when “the code stays on our hardware” is non-negotiable.
Tabnine — privacy-first for enterprise
Tabnine targets the same privacy-conscious crowd but at the enterprise end. It supports SaaS, VPC, on-premises, and fully air-gapped deployment, with zero code retention and the compliance stack (SOC 2, ISO 27001, GDPR). Pricing: Code Assistant at $39/user/month billed annually, an Agentic Platform tier at $59/user/month, and Enterprise with custom pricing for on-prem, air-gapping, and bring-your-own-LLM.
One cost people forget: if you self-host the models, the per-seat fee only covers the license. You still pay for GPU compute — realistically $500–2,000+/month in cloud GPU or equivalent on-prem hardware on top of the seats. Budget for it. Tabnine fits regulated industries and air-gapped environments where Copilot-in-the-cloud is simply off the table.
Aider — the terminal holdout
If you live in the terminal, Aider is an open-source CLI pair programmer where you bring your own API key (or a local model) and pay only for what the underlying model costs. No subscription, no vendor account to lose. It won’t give you ghost-text completions in your editor, but for “describe a change, let it edit across files, review the git diff” it’s excellent. Niche, but the people who love it really love it.
A quick decision matrix
No single right answer here — it depends on what you actually are:
- AWS-heavy team that liked the vendor path → Kiro. The integration story and frontier-model access are strongest here, and you’re already in the ecosystem.
- Polyglot startup that just wants it to work → GitHub Copilot. Best editor support, broadest models, lowest migration risk. Just model the June 1 usage-based billing change first.
- Privacy-constrained or regulated enterprise → Tabnine for managed air-gapping, or Continue.dev if you want open-source and don’t mind running the infra yourself.
- Solo dev or cost-sensitive → Gemini Code Assist’s free tier (with the Antigravity caveat) or Aider with your own key.
The migration checklist
However you jump, do these before April 30, 2027 — and honestly, well before, because the second half of that window will be crowded with everyone else procrastinating:
- Inventory what you actually used. Inline completions only? Chat? Agents? The CLI? Your answer narrows the field fast.
- Export anything custom. Saved prompts, custom rules, agent configs. Don’t assume they’ll still be reachable on April 30, 2027.
- Swap the IDE plugin and run a pilot. Install the replacement alongside Q on one repo for a sprint. Don’t rip Q out on day one — they coexist until the cutoff.
- Re-test your CLI and CI workflows. If you scripted Q into anything automated, that’s the part that breaks silently.
- Right-size the plan after two weeks of real usage, not on day one. Usage-based billing (Copilot) and per-interaction caps (Kiro Free) mean your first guess about cost is probably wrong.
- Pick a tool with a roadmap you trust. You’re migrating because one got sunset. Two of the obvious replacements — Gemini Code Assist and, well, Q itself — are mid-transition right now. Factor longevity in.
You’ve got twelve months. The teams that treat this as a chance to actually reconsider what they want from an AI assistant will come out better than the ones who reflexively click “migrate to Kiro” and move on. Spend an afternoon running two of these side by side on a repo you know well — that’s worth more than any comparison table, including this one.
Dates and pricing are accurate as of May 2026 and are moving targets in this category. Check the AWS announcement and each vendor’s pricing page before you commit.