Replit has unveiled Agent 3, the latest version of its AI-powered Agent, designed to give developers more time back while ensuring smoother, more reliable app creation. Compared to Agent V2, this upgrade represents a significant leap forward in autonomy, efficiency, and flexibility.

Smarter App Testing with Real Browser Interaction

One of the standout features in Agent 3 is App Testing. The Agent now tests the apps it builds using an actual browser clicking through buttons, forms, APIs, and data sources. By doing so, it ensures that every part of your app functions properly.

When enabled, the Agent not only tests but also fixes issues automatically, offering summaries of what it checked and what it resolved. For apps using Replit Auth, it can even test the login flow.

For those working on complex projects, the Max Autonomy (Beta) option allows Agent 3 to run longer sessions (up to 200 minutes), manage extended task lists, and monitor its own progress with minimal supervision.

Agents That Build Other Agents

Agent 3 introduces the ability to create other agents and automations. With just natural language input, you can now design bots and workflows that integrate seamlessly with tools like Notion, Linear, Slack, Dropbox, and SharePoint.

Examples include:

1. A Slack bot that answers codebase questions by pulling from GitHub.
2. A Telegram bot that schedules appointments on Outlook Calendar.
3. Automated daily task summaries delivered via email.
4. Meeting prep emails that gather guest details, summarize with AI, and save notes to Google Drive.

Instead of juggling API keys, Agent 3 guides you through a simple connection flow, ensuring a smooth setup.

Extended Runtime and Flexible Build Modes

Autonomy is at the core of Agent 3. It can now run independently for up to 200 minutes, handling testing, fixing, and building without frequent user input. Progress can be tracked in real time whether on your homepage or your phone.

For creators who prefer flexibility, the new app creation flow offers two modes:

1. Frontend-only mode for quick prototyping and design iteration.
2. Full-stack mode for end-to-end app building, with the option to add a backend later.

This means you can move at the pace your project requires.

Opening the Door to More Possibilities

By combining extended autonomy, built-in testing, and workflow automation, Agent 3 makes app development more fluid and adaptive. Whether you’re experimenting with a quick prototype or setting up a complex automation system, this Agent is designed to meet your needs without unnecessary complexity.

The possibilities of what you can build, automate, and test are broader than ever.

Try Agent 3 today on Replit and experience what autonomy can add to your workflow.