Replit August 21, 2026 Changelog
What changed in Replit & Replit Agent August 21, 2026, released 2026-08-21 — explained in plain English.
TLDR
- You can now chat and build for free using a new "Free Mode" without burning through your paid credits.
- You can turn casual chats into full projects, automate repeating tasks, and guide the AI while it works.
- There are new ways to share AI skills from GitHub and run advanced security tests on your apps.
- This update is all about making the AI helper easier to use and more flexible. It gives you more ways to build for free, keeps your conversations organized, and adds tools to automate tasks and secure your code. This matters to anyone who uses Replit to code, from solo students to large enterprise teams.
Want a heads-up when Replit updates?
We email you the moment a new Replit version drops — plus plain-English release notes like these. No spam.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
What's new
- Free Mode: You can now chat, brainstorm, and build without using up your paid Power Mode or Max Mode credits. Free Mode automatically picks the best AI model (a type of AI brain) for you, and your limits depend on your Replit plan.
- Conversations: You can start with a private Conversation to ask questions or attach files. When the idea gets big enough, you can tell Replit to turn that chat into a full Project, bringing all your context and files with it.
- Routines: You can schedule Replit to automatically repeat tasks inside a Conversation on a schedule. Each time the task runs, the results drop right back into the chat for you to review.
- Import Skills from GitHub: You can now pull custom AI "Skills" (instructions or tools for the AI) straight from a public GitHub code repository. You can also set permissions so only certain team members can use specific skills.
- Enterprise Model Controls: Admins (the people managing a company's Replit account) can now choose exactly which AI models their teams are allowed to use in different Workspaces.
- Black-box Pen Tests: You can now run a Level 3 security scan on your app. This combines free checks on your app's outside packages, an AI review of your source code, and a "black-box" test (where an outside tool tries to hack the live preview of your app to find vulnerabilities).
Improvements
- Steer Agent while working: You can now send a follow-up message to the AI while it is still working on a task. This adds new information to what it is currently doing. If you prefer it to wait and do things one by one (called Queue), you can change that in your settings or just press
Cmd+Enter(Mac) orCtrl+Enter(Windows/Linux) for a single message.
Fixes
*No bugs or fixes were mentioned in this update.*
Breaking changes
No breaking changes in this release.
This is our plain-language summary. Read the complete, official notes on the Replit official changelog ↗.
Google Calendar