We’re excited to introduce Agent 3—our most advanced and autonomous Agent yet. Compared to Agent V2, it is a major leap forward. It is 10x more autonomous, with the ability to periodically test your app in the browser and automatically fix issues using our proprietary testing system—3x faster and 10x more cost-effective than Computer Use models. Even better, Agent 3 can now generate other agents and automations to streamline your workflows.
What’s New
1. App Testing: Agent tests the apps it builds (using an actual browser)
Agent 3 now tests and fixes the app it is building, constantly improving your app behind the scenes. We are launching two different options here, depending on your needs:
- App Testing is on by default, but can be turned off within the Agent Tools section of the input box within the workspace.
- Max Autonomy (Beta) lets Agent run for longer and manage itself—ideal for complex, extended tasks with well-defined goals.
When App Testing is on, the Agent will periodically decide to test your application. You’ll be able to see a browser preview within the Agent pane, showing the Agent’s cursor as it clicks around the app. It checks buttons, forms, APIs, data sources etc — ensuring every part of your app works perfectly. The Agent will reply back with a summary of its tests, and fix any issues that it detected. Agent can even automatically log into applications built with Replit Auth to test the user login flow.
Note: the Agent will not test the App created after 100% of user messages, instead it will decide to test the App when it thinks enough has changed to deem it necessary.
With Max autonomy, Agent can run for extended periods of time (200 minutes or more) with minimal supervision, exhibiting the following behaviors:
- Extended sessions: Agent runs longer without requiring user input.
- Task management: Agent able to create and work through longer task lists.
- Self-supervision: Agent monitors its own progress during the session.
2. Agent can now build other Agents & Automations
For the first time ever, Agent 3 can build other agents and automations. That means you can automate complex and repetitive workflows using natural language. Just select the “Agents & Automations” option from the homepage to get started.
With this option, you can build things like Telegram bots, Slack agents, and time-based automations (e.g. daily reminders). Just like building normal apps, you’ll describe what you want in words. However, unlike most Replit apps there’s no preview — you’ll deploy Agents & Automations to use them in another surface via an integration (e.g. in Telegram, Slack or Email).
When building an Agent or Automation, on one side of the workspace you’ll see the Agent pane, just like other types of Replit apps. On the other side, instead of a preview you’ll see an admin dashboard where you can test your chatbot or workflow automation.
You can also build automations with apps like Notion, Linear, Dropbox, Sharepoint etc. For example, when you ask an Agent to build something that involves Notion, it now guides you through a seamless connection flow. Instead of manually hunting down and pasting API keys, the Agent surfaces a simple UI flow to connect with Notion. Once finished, the Agent securely receives the credentials it needs and continues building automatically.
Automations made possible
With Agent 3’s ability to build Agents and Automations, a whole new range of possibilities opens up. For example, you can create automations like:
- Slack bot that answers questions about your codebase by integrating with GitHub.
- Slack bot that retrieves and queries information from your Notion database.
- Automated daily email summarizing your tasks from Linear.
- Automated meeting prep email: sent 20 minutes before each external meeting, it searches the web for details about your guest and their company, summarizes key points with AI, and saves the notes to Google Drive.
- Telegram bot that can schedule appointments directly on your Outlook Calendar.
3. Longer Autonomous Runtime
Agent 3 runs on its own for up to 200 minutes, handling full tasks autonomously—building, testing and fixing your app, with minimal manual oversight, giving you hours of time back. You can track your project’s progress, in real time, on your home page on the web or live monitoring on your phone.
4. New App Creation Flow
While Agent is now more powerful, we know you don’t always need full-stack capabilities—sometimes you just want to prototype quickly or iterate on designs before deciding to build a full app. That’s why we’re launching a new app creation flow that gives you flexibility: either build a full-stack app right away, or start with frontend-only for rapid prototyping and design iteration and add a backend later if you’d like.
You can select the right mode to ensure Agent always works in the mode that best fits your goals.
Autonomy for All
This is just the beginning—Agent will keep getting more powerful as we climb the autonomy curve, with more integrations, trigger-based automations, and other updates to make building anything on Replit easier than ever.
Agent 3 is available for all free and paid users, and we can’t wait to see what you build with it. Try Agent 3 today!
