Changelog

New

• Claude Opus 4.5 is Now Default - As the creator of Playcode, I believe in giving you the best experience from the start. The AI now uses Claude Opus 4.5 by default, the most capable model available. It costs more per message, but the quality difference is immediately visible. Magic should feel like magic.
• Cleaner Chat Interface - Removed the model selector from the chat toolbar for a simpler experience. Power users can still change models via Menu → Preferences → AI Model.

Improvements

• Reduced AI System Prompt - Optimized the AI system prompt by ~60%, making each request slightly cheaper.

Fixes

• Chat Character Escaping - Fixed an issue where HTML entities and escape sequences were displayed in the chat preview instead of actual characters.
• Preview URL Dropdown - The HTML page selector in the preview toolbar now stays open when clicking, allowing you to actually select a page.

Improvements

• AI Import Now Works with Full Responses - Copy the entire AI response including explanations and markdown. We automatically extract the JSON code block and apply changes. No more "Failed to parse" errors!
• Helpful Tooltips - Hover over Copy, Paste, and Apply buttons to see step-by-step instructions on how to use the ChatGPT/Gemini workflow.
• Better Error Messages - When something goes wrong, you now get actionable guidance on how to fix it.

New

• ChatGPT/Gemini Import/Export Buttons - Copy your project files with one click, paste into your AI, then paste the response back and apply changes instantly. Includes Undo/Redo to rollback if something breaks.
• Inspiring Placeholder - The AI input now shows animated examples of what you can build: customer portals, admin dashboards, landing pages, and more.

Improvements

• Consistent Design - All primary action buttons now use the signature orange gradient for a unified look across popups and notifications.
• Pro Badge Styling - Updated Pro badges to match the marketing design with the correct gradient colors.

Improvements

• Truly Empty Templates - Empty JavaScript, TypeScript, HTML, and CSS templates are now minimal. Just the bare essentials with a single comment to get you started from scratch.

Fixes

• Live View Delay - Changing the delay setting now applies immediately. Previously, adjusting the delay in preferences had no effect until page reload.

Changes

• Free tier adjustments - To keep Playcode sustainable and continue offering a great experience, we've had to reintroduce some coding limits for free accounts. Pro users continue to enjoy unlimited lines of code, along with all other premium features. We appreciate your understanding and support!

New

• GPT-5.2 Model - OpenAI's most capable model is now available. Advanced reasoning and planning capabilities for complex coding tasks.

Improvements

• Cleaner Read-Only Presentations: When sharing a read-only project, only the owner's cursor is now visible. Viewer cursors are hidden, preventing visual chaos when presenting to large groups like classrooms with 40+ students.

Fixes

• Binary Assets in NPM Packages: Fixed support for npm packages that include binary assets like fonts, images, audio, and video files. Packages like @expo/vector-icons, icon libraries, and media packages now work correctly.

Major Update: Unlimited Free Coding!

We've removed all code limits. Playcode is now completely free to use with unlimited coding in the playground.

New

• Unlimited Free Coding - We've completely removed the 8-line code limit! Code as much as you want, experiment freely, build anything - no restrictions, no timers, no paywalls.
• Simplified Pricing - One simple plan: Playcode Pro at $9.99/month or $99/year (save $20 annually). No confusing tiers, no feature gates.
• Embedded Stripe Checkout - Upgrade without leaving Playcode. Supports all payment methods: cards, Apple Pay, Google Pay, bank accounts. 3D Secure handled automatically.
• Instant Cancellation Option - New "Cancel immediately" option for those who need to downgrade right away (standard cancellation still available).
• Plan Switching in Billing Portal - Switch between monthly and yearly billing anytime. Get prorated credits when upgrading to annual.
• Collaboration Limits - Free accounts support 1 collaborator (perfect for pair programming and 1-on-1 interviews). Pro gets unlimited collaborators for team sessions.
• Project Export (Pro) - Download your projects as ZIP files with source code and production build. Pro-only feature.

Improvements

• Grandfathering Policy - Already have 100 projects? Keep them all! The 3-project limit only applies to NEW saves. We don't punish early adopters.
• Smarter Upgrade Prompts - Hit a limit? See exactly what you have (your 3 projects listed) and what you'd get with Pro. No more confusion.
• Stripe Billing Portal Integration - Full control over your subscription: download invoices, update cards, add tax IDs, switch plans - all without contacting support.
• Real-time Webhook Sync - Cancel in Stripe? Reactivates in Playcode instantly. Switch plans? Updates immediately. Everything stays in perfect sync.
• Project Counter in Dashboard - Free users see "2/3 projects" right in the header. Always know where you stand.
• Reactivation Support - Changed your mind? "Don't Cancel" button brings back your subscription instantly, or use the billing portal.

Fair Downgrade Policy

If you cancel, here's what happens:

• Keep all existing projects - Even if you have 100, they're safe (grandfathered)
• Private → Public - Your private projects become public
• Custom domains removed - Sites get random number subdomains
• Storage: 1GB → 4MB - Large assets over 4MB will be deleted
• Keep using Playcode - Continue coding, just can't save new projects over limit

Questions? Check our billing help articles or [chat with us](#).

Fixes

• Chat Connection Stability: Fixed multiple bugs causing the AI chat to disconnect unexpectedly. The chat is now much more stable and reliable.
• Agent Task Interruption: Fixed an issue where the AI agent would randomly stop working on tasks without being prompted to do so.
• GPT-5.1 Codex Pricing: Fixed incorrect cost calculation for GPT-5.1 Codex model usage.