What’s new in Playcode

156 updates in the last 90 days.

New features, improvements and fixes, newest first.

RSSJSON FeedMarkdown for agents

Our rocket on its pad beside a tall board of update cards: the builder pins a fresh card at the top, Play fits a new part to the rocket, and two passing suits stop to stare.

November 2026

NewBuild

GPT-6.2 Sol and GPT-6.2 Astra join the model picker, and your apps can call them

GPT-6.2 Sol and GPT-6.2 Astra, OpenAI's newest models, replace GPT-6.1 Sol and GPT-6 Astra in the model picker: GPT-6.2 Sol on the Standard level, GPT-6.2 Astra on Ultra. Both hold a 1M-token context window and read images.

  • Same cost dots. GPT-6.2 Sol shows two, GPT-6.2 Astra three, as the models they replace.
  • For your apps too. Apps you host on Playcode can call both models through Playcode AI Gateway, and choose how much they reason, from low to maximum effort.
  • See them drawn. Both models join the MacBook SVG benchmark, at every reasoning level.

Read the docs

NewAdd to your app

Qwen4-Max for your apps

Apps you host on Playcode can now call Qwen4-Max, Alibaba's new flagship model for coding, agent work and long documents, through Playcode AI Gateway. It holds a 1M-token context window and reads images.

  • Choose how much it reasons, or turn reasoning off, for each request.
  • See it drawn. Qwen4-Max joins the MacBook SVG benchmark, at every reasoning level.

Read the docs

ImprovedAccount and billing

Billing shows a plan payment that did not go through

When a renewal, or a move to a bigger plan, cannot be charged, Billing & plans now says so: "Your plan payment failed", with the day, the amount and the day Stripe tries again.

  • Update card opens Stripe's billing portal, where you change the card and pay.
  • The agent knows too. Ask whether your last plan payment went through, and it reads the same answer.
  • Credits follow the payment. A move to a bigger plan adds its credits once its payment goes through, for what you paid, up to the difference between the two plans. Until a failed payment goes through, Billing shows no date for your next monthly credits.

Read the docs

Try it: Did my last plan payment go through?

NewAdd to your app

Gemini 4 Carbon and Gemini 4 Hydrogen for your apps

Apps you host on Playcode can now call Gemini 4 Carbon, Google's new model for its hardest, longest tasks, and Gemini 4 Hydrogen, its fast model for everyday coding and agent work, through Playcode AI Gateway. Both hold a 1M-token context window, read images and PDFs, and run on Google Cloud in the EU.

  • Choose how much they reason, from low to high, for each request.
  • See them drawn. Both models join the MacBook SVG benchmark, at every reasoning level.

Read the docs

FixedPublish

The agent and the terminal now see environment changes right away

When you change an environment variable in a Cloud project, in Settings > Environment or through the agent, your running app gets the new value right away. Now the agent's own commands and the Terminal get it too, with no restart.

Before, they kept the values from when your app's cloud computer last started, which could be weeks earlier. So when the agent checked a variable or ran a script, it could see a value you had already changed or removed, or miss one you had just added.

A terminal that was already open keeps its old values: open a new one to get the new ones.

FixedBuild

Background tasks carry on through Playcode updates

When Playcode AI hands a build to a background task, the task could end with a wrong "server restart" error in two cases:

  • A task that was working while Playcode updated could stop about ten minutes later instead of carrying on.
  • When a task got a change just as it finished a step, its real end could be lost. Even "the credits ran out" was reported as a restart.

Now a task picks up where it was after an update, and it always reports how it really ended. Your work in those tasks was always saved.

FixedPublish

Domain changes in the chat ask in your language

When you ask Playcode AI to change a domain, it asks you to confirm in the chat first. For five changes - bringing a domain here from another registrar, moving it to another workspace, serving it straight from Playcode, disconnecting it from a site and removing its settings from Playcode - that card showed an English sentence in every language, and named a workspace or a site by an internal ID.

Now each of these cards says what happens, in your language:

  • Bring a domain here: the price, what it renews at after the move, the card it charges, and how long the move takes. The button states the price, and the move charges that price or nothing.
  • Move to another workspace: the workspace and the site by their names, and who pays the next renewal.
  • Straight from Playcode: what changes for visitors where the delivery network is blocked, the one setting to add when the address's settings live at another provider, and the trade.
  • Disconnect and remove from Playcode: which site stops answering, what stays, and what goes offline.

The reply your click sends names the change too.

FixedPublish

A failed publish now says what stopped it

When a publish failed, Playcode said "The build failed" and asked you to check your code, even when the connection dropped or something on our side stopped it. Playcode AI then looked for a mistake in code that was fine.

Now the message says what happened:

  • The connection: "Could not reach Playcode". Check your internet and publish again. If it happens twice in a row, your network may be blocking us: try another one, such as mobile data.
  • Our side: "We could not finish the publish". Publish again, and reload the page if it happens again.
  • Your code: "The build failed", with the error, as before.

Ask the agent to fix it now tells Playcode AI the same thing, so it tells you what to do instead of changing code that works.

If your package.json asks for a package version that does not exist, the error now names the package and the version.

NewBuild

Claude Fable 5.2 joins Ultra, and your apps can call it

Claude Fable 5.2, Anthropic's newest Fable model, joins the Ultra level: open its model list and pick it under Claude Fable 5, which stays the default. It holds a 1M-token context window, reads images, and shows a summary of its thinking as it works.

  • See how it draws. Claude Fable 5.2 has its own row in the MacBook SVG benchmark, at every effort level from low to max.
  • For your apps too. Apps you host on Playcode can call Claude Fable 5.2 through Playcode AI Gateway.

Read the docs

FixedBuild

In a long chat, Playcode AI keeps the files you sent

When a chat grows long, Playcode AI folds its older messages into a short summary and keeps working. The screenshots and files you sent before that point left with those messages, so Playcode AI could ask you to send the same screenshots again.

Now Playcode AI keeps the links to the files you sent and notes what each one is for. When the work needs a screenshot again, it opens it from its link instead of asking you.

FixedPublish

A stuck upload no longer fails a checkpoint or a publish

On a Cloud app, a checkpoint or the backup before a publish could fail when one upload to storage got stuck. Nothing was lost, but you or Playcode AI had to try again.

Now a stuck upload also goes another way after a short wait, so the checkpoint or the publish carries on.

ImprovedPublish

A sleeping Cloud app in Finland now wakes in about 0.4 seconds

A Cloud app sleeps when nobody visits, and the next visitor wakes it. In Finland, that first answer took about 1.4 seconds a month ago. Now it takes about 0.4 seconds, with the app's memory right where it left it.

  • Nothing to change. Every Cloud app gets this, on any plan.
  • What the number covers. It is the server's side of the wake, measured on a small test app. A slow first page in your own app still adds its own time.

FixedBuild

The preview no longer says it could not start when it did

A few seconds after you started a new project, the preview could show "Preview couldn't start", with a Try again button. Your page was fine behind that card. It looked like a slow connection, but it was our bug.

Now the card appears only when the preview really did not start.

  • On a new project, Playcode AI now waits for the preview and checks your page. It no longer stops at the build because the preview seemed closed.
  • If the preview really cannot start, Playcode AI says so at once, and you get the same card with Try again.
  • An image that has no source yet, like a photo a script fills in later, no longer reads as a file that did not load.
  • A page Playcode AI has just created now opens as itself. The preview could show the previous page instead and keep it through later edits, until you reloaded.

NewBuild

Animated GIFs, SVG logos, and pictures at about half the credits

Playcode can now make animated GIFs for your site: a waving mascot, a loader, an animated sticker. Describe the motion, or give it a picture you already have - your logo or your mascot - and ask it to bring the picture to life. It keeps your picture and moves only what you ask.

It can also draw SVG logos and icons: vector files you can edit and scale to any size.

Pictures now cost about half the credits they did before today.

GPT Image 2.6 joins as an option when you want a sharper, higher-contrast look. The same picture uses about three times the credits on 2.6, so GPT Image 2.5 stays the default.

Image generation is part of the paid plans.

FixedPublish

Sites on the Free plan can use the camera and the microphone

A page that used the camera or the microphone worked when you opened it alone. On a site published on the Free plan, the browser refused it at once and never asked the visitor. The page's code was not at fault, so Playcode AI could not fix it.

Now the browser asks the visitor, as on any other site.

  • On a site published on the Free plan, a page can use the camera, the microphone, screen sharing, the clipboard, the location and the motion sensors. Fullscreen stays off.
  • In the preview of a static site, a page can use fullscreen, copy to the clipboard and read the motion sensors. For the camera, the microphone, screen sharing, pasting and the location, choose Open preview in a new window, or publish the site.

Sites on paid plans were not affected.

ImprovedBuild

Playcode AI can check your page in your own browser

Playcode AI could change your code without seeing the result, and it could say it was done on a page that did not work. You were the one who told it.

Now it can look at the preview in your browser while it works, in static sites and in Cloud apps.

  • It sees what you see: build errors, errors on the page, failed requests and a blank page. That includes an error that appears only after you click something, and the page keeps its state.
  • If it cannot look, for example when your tab is closed, it knows that it did not check.

RemovedBuild

Gemini 4 Argon's introductory price has ended

Gemini 4 Argon now runs at Google's standard price: $4 in / $20 out per million tokens instead of $2 / $10. The same work on Argon uses twice the credits it did before today.

When Argon joined the Standard tier on October 14, we said its introductory price would run through November 30. It ended earlier than we said.

If you chose Argon for its price, GPT-6.1 Sol and Claude Sonnet 5.5 on the same tier cost $2 in / $10 out per million tokens.

GPT-5.6 Sol, which saved chats and apps can still use, also returns to its standard price: $5 in / $30 out instead of $4 / $20.

NewBuild

A long chat tells you what the next message costs

After an hour without messages, a level switch, a new plan or a move to Cloud, the next message in a long chat reads the whole chat again, and that costs more. Now the chat says so before you send: above the message box you see Next message: about N credits extra and why.

  • New chat starts the next task in a clean chat at the same level. Your draft moves with you.
  • Compact chat keeps your recent messages and a summary of older ones, so the next message reads a smaller chat. It shows its cost, and it is offered only when it saves at least 5 credits. Small early details can get lost.
  • Send as is: the AI sees the whole chat. Only the next message costs extra.

Read the docs

ImprovedBuild

Language templates open as a playground: code, preview, console

Open a JavaScript, TypeScript, React, Python, SQL or any other language template and you get a clean playground: the file tabs and the editor on the left, the live preview and the console on the right.

The AI Chat is one step away: open the project menu and choose AI Chat in the View group, or press Ctrl+Shift+C (⌘⇧C on a Mac). On a phone, a language template opens on the preview, and the Chat button opens the chat.

Website and business templates keep opening chat-first: describe what you want, and the preview updates as you go.

FixedBuild

Stopping a task now stops it

When you stopped a background task in the tasks panel, or Playcode AI stopped one, the panel said Cancelled. But the task could go on working until it finished: it changed files and used credits.

Now a stopped task stops working.

FixedBuild

A short model failure no longer stops the work

Sometimes the AI model fails for a moment in the middle of an answer. Playcode AI then stopped: the chat showed an error, and a background task ended without its report, even when its changes were already saved. You had to write again to restart the work.

Now Playcode AI tries that step again and goes on.

ImprovedPublish

New Cloud apps: you become the admin with a link from Playcode

In a new Cloud app, signing up no longer makes anyone the admin, not even the first account. You become the admin through an owner link that Playcode gives you in the chat: open it while signed in to your app, and your account owns the app. The link works once, for 30 minutes. Ask in the chat for a new one at any time.

  • After you turn a static site into a full-stack app, Playcode's next reply tells you how to reach its admin at /dashboard, and offers a small Admin link in your landing's footer.
  • Your live app has its own sign-in. The first publish signs everyone out of the live app and clears its admins, so an account from the development copy is not an admin there. Playcode gives you a second link to claim the live app.
  • Sign-in protects your users. After several wrong passwords, the form makes the next try wait. A forgot-password page sends a reset link that works once.

Try it: Give me the owner link for my app.

FixedPublish

Videos and images with #t= or ?v= in their link now go live

A video tag like <video src="clip.mp4#t=1"> shows the first frame as a preview, and an image link can carry a version, like logo.png?v=2. Playcode left such a link as you wrote it, so your published site could not find the file and showed nothing.

Now the build finds the file and keeps the #t=1 or ?v=2 on its link, so the video or image shows on your live site.

ImprovedBuild

Playcode AI shrinks videos up to 150 MB

A video straight from a phone is often over 100 MB. Playcode AI could look at, cut or shrink a video only up to 80 MB, so a bigger one went into your site as it was, and your visitors waited for it to load.

Now it works with every video you can attach to the chat, up to 150 MB. Attach the video and ask for a smaller version for your website.

Try it: Make this video smaller for my website.

FixedBuild

Playcode AI sees 3D pages in its screenshots

When Playcode AI took a screenshot of a 3D page, such as a game built with three.js, it saw the page's "WebGL is not supported" message instead of the scene, and could take that message for a bug in your page.

Now its screenshots show the 3D scene, and so do the thumbnails of 3D projects.

FixedPublish

Publish works while the preview is still loading

If you pressed Publish right after you opened a project, while the preview was still loading, your live site could come out blank. Publishing again fixed it.

Now Publish waits for the preview to finish loading, then builds your site. If a built page would still be broken, Playcode stops the publish and says so, and your live site stays as it was.

FixedBuild

A delete in the chat names what goes, in your language

When you ask Playcode AI to delete something - a project, a chat, a checkpoint, a saved card - it asks you to confirm in the chat first. That card showed a long internal ID and an English sentence, so with several projects you could not tell which one would go.

Now the card names it the way you know it: the project's title, the card's brand and last four digits, the checkpoint's time. It says what goes with it and that this cannot be undone, in your language. The reply your click sends names it too.

NewAdd to your app

Your Cloud app has its own Assistant

New Cloud apps now start with an Assistant for the app's admins, at /assistant, second in the admin sidebar. Ask it about your business in plain words, such as "How many new leads came in today?", and it answers from your app's own data.

  • It works through your app's operations. Every feature of a new Cloud app is an operation: its pages and the Assistant call the same route, with the same checks. When Playcode adds a feature to your app, it adds it as an operation, so the Assistant can use it.
  • It shows where each answer came from: the operation, the filter and the result, under the answer.
  • Every change waits for you. The Assistant proposes a change on a card that shows the exact change and the record as it is now. Nothing changes until you press Apply. Before a delete, the card says what goes.
  • What customers write stays data. A lead's message that says "ignore your instructions" cannot make the Assistant act on its own, and a card says when its proposal came after reading what a customer wrote.
  • Activity lists every change to your app's data: what changed, who made it, when, and whether it came from a page or the Assistant.
  • It uses AI credits only when an admin asks it something. Opening the page costs nothing.

Try it: Add an operation to my app so its Assistant can cancel a booking.

NewDevelopers

Playcode SDK 0.8: your app's features as operations

@playcode/sdk 0.8 (npm) is out. New Cloud apps start on it. In an older app, ask the agent to update the SDK: it reads the upgrade steps in the SDK's changelog first.

  • Your app's operations. @playcode/sdk/operations describes each feature once, as a route with a description. Your pages, an assistant with four tools, and an MCP client with an app token call the same route through one executor. A lint checks every route, and operationsContract proves your app keeps its half: Your app's operations.
  • Each change waits for a person. operationCard() builds the card from the exact arguments and your app's own reads, never from the model's words. Customer-written text reaches the model inside marked blocks, and one run asks a person about at most 3 changes and 3 sends.
  • An approval check sees the run. Its second argument lists the calls the run already put to a person. Past your own line it answers { cap }, and the run ends with the new finish: "cap".
  • A run keeps its session. Pass session to agent.handler(), run() or start(). Every tool call reads it, so a sign-out stops the run at its next call.
  • What each agent spent. A named agent's model calls carry its name, so your app's AI usage reads by agent.
  • Results for your page. A tool result can carry data that your page shows and no model reads.

Before you upgrade: a store with its own tables needs two new nullable columns, session on runs and sources on threads, and agentStoreContract checks both. A page that switches on finish adds the cap case.

Read the docs

Try it: Update my app's Playcode SDK to the latest version.

FixedBuild

Try again restarts a preview that could not start

Sometimes the preview cannot start, most often on a slow or unstable connection. It then shows Preview couldn't start with a Try again button.

Try again used to reuse the part that had failed. The preview then said "Your page can't be shown yet" and blamed your code, so Fix it sent Playcode AI to change files that had nothing wrong with them. Only a reload of the page helped.

Now Try again starts the preview fresh, and your page shows as soon as it loads.