Replit Agent

Agent uses powerful AI technology to transform your ideas into applications and seamlessly add new features by describing what you want.

​

Features

Chat with Agent in your Replit App to receive the following support:

• Create full-stack apps from scratch
• Add advanced features and integrate complex APIs
• Design, create, and modify database structures
• Streamline environment setup and dependency management

​

Build Modes

When you submit your initial prompt to create an app, Agent creates a feature list based on your request. You’ll then see the question “How do you want to continue?” with two build options: You can also choose to “continue planning” if you want to refine the feature list before building begins.

​

Start with a design

Perfect for seeing visual results first:

• Quick design phase: Agent creates a clickable front-end prototype in ~3 minutes
• Visual-first approach: See how your app will look and feel before building functionality
• Flexible continuation: Press “Build functionality” to have Agent complete a full working version of your app
• User-controlled pacing: You decide when to move from visual prototyping to a finished application

After the design phase, you’ll see options to continue:

​

Build the full app

Ideal for comprehensive application development:

• Complete functionality: Agent builds out the full functionality of your app from the start
• Initial build: Agent creates an initial working application in ~10 minutes
• Comprehensive scope: Full-stack development with frontend, backend, and integrations

After the initial build phase, you can:

• Review the full task list: See exactly what Agent plans to build next
• Accept or modify: Approve the plan or request changes before full implementation
• Complete the build (New in Agent 3): Let Agent finish the comprehensive development (can run for up to 200 minutes)

After creating an MVP, you’ll see options to continue building:

​

Autonomous Features (New in Agent 3)

Agent 3 introduces powerful autonomous capabilities that enable extended, self-supervised development with minimal human intervention.

​

App Testing

Automated browser testing - Agent tests itself using an actual browser, navigating through your application like a real user would. Key capabilities:

• Real user simulation: Agent clicks through your app, testing functionality and user workflows
• Automatic issue detection: Identifies problems and fixes them during development
• Visual feedback: Provides video replays of testing sessions for review
• Intelligent timing: Agent decides when testing would be most valuable

Usage: Toggle “App testing” in the Agent Tools panel. When enabled, Agent intelligently decides when testing adds the most value. Learn more about App Testing capabilities and troubleshooting.

​

Max Autonomy

Extended autonomous development - Agent works for much longer periods (up to 200 minutes) with minimal supervision. Key benefits:

• Extended work sessions: Agent can work much longer without requiring input
• Longer-tasklist handling: Creates much longer task lists to complete more functionality
• Reduced supervision: Agent will supervise itself, so you don’t have to (runs up to 200 minutes)

Usage: Toggle “Max autonomy” in the Agent Tools panel based on your comfort level with autonomous development. This feature is currently in Beta.

​

Agents & Automations

Beyond traditional apps - Build intelligent agents, chatbots, and automated workflows that interact with external services. Supported use cases:

• Slack Agents: Intelligent Slackbots for research, Q&A, and task automation
• Telegram Bots: Customer service, scheduling, and entertainment bots
• Timed Automations: Scheduled workflows for reports, summaries, and monitoring

Getting started: Select “Agents & Automations” from the Replit homepage app type selector to begin building automated systems. Learn more about Agents & Automations use cases and setup.

​

Usage

You can access Agent in your Replit App workspace by selecting the Agent tool.

How to access Agent

From the Agent tool, you can perform the following actions:

• Chat: Describe your app or feature in the text area and respond to Agent’s follow-up questions
• Add detailed requests: Upload files or import content from URLs to give Agent more information
• Manage conversations: Select New chat or select previous conversations from the menu icon to start or resume a conversation
• Track usage: Select the usage icon at the top right to view your billing page
• Reverse changes: Undo changes made by Agent by selecting Rollback to here

​

Chat prompts

To communicate your request to Agent, enter a prompt in the text area. A prompt is a natural language instruction that describes the task you want Agent to perform. In addition to text, you can include the following data in your prompt:

• File attachments: Drag a file into the text area or select the paperclip icon
• Web content: Include text data from a webpage by entering a URL and selecting Copy page content
• Webpage screenshots: Include a screenshot of a webpage by entering a URL and selecting Take screenshot

Agent operates in three distinct modes to support different types of work:

• Build mode (default): Agent writes code, modifies files, and implements features directly in your project
• Plan mode: Brainstorm ideas, plan development work, and collaborate on project architecture without modifying code. Learn more about Plan mode
• Edit mode: Make targeted changes to specific files or code sections with precise control

​

Checkpoints

A checkpoint is a comprehensive snapshot of your Replit App that captures completed work from Agent, including workspace contents, AI conversation context, and connected databases. Agent creates meaningful checkpoints when it finishes implementing your request, ensuring you only pay for completed functionality.

​

How checkpoints work with effort-based pricing

Agent’s new effort-based pricing creates checkpoints that reflect the actual work performed:

• One checkpoint per request: Agent bundles all work for your request into a single, meaningful checkpoint
• Variable pricing: Simple changes cost less, while complex builds cost more based on the effort required
• Transparent costs: Each checkpoint shows exactly what you’re paying for that completed work
• No intermediate billing: You’re not charged for Agent’s thinking or planning process

Each checkpoint lets you perform the following actions:

• Rollback: Undo changes made by Agent and return to the state of the previous checkpoint, including workspace contents, AI memory, and connected databases
• Track usage: Hover over the usage icon in a checkpoint to view the amount billed for that checkpoint

​

Advanced capabilities

For complex projects, you can enable additional Agent capabilities through toggles in the chat toolbar: Extended Thinking: Instructs Agent to analyze your request more thoroughly, considering multiple approaches and edge cases. Best for architectural decisions or when you need comprehensive solutions. High Power mode: Uses more sophisticated AI models for advanced problem-solving. Ideal for complex integrations, performance optimizations, or unfamiliar technologies. Learn more about when and how to use these advanced features in the Dynamic Intelligence documentation.

​

Progress tab

The Progress tab shows a record of Agent’s actions in real-time. Access it through the Tools dock or search bar to monitor Agent’s progress and navigate directly to relevant files. Key features:

• Real-time updates: Monitor Agent’s progress with live activity feed
• Chronological history: Review Agent’s activities to understand changes and troubleshoot issues
• File navigation: Click links to jump directly to files Agent is working on

​

Getting the most from Agent

​

Best practices for prompts

• Be specific: Describe exactly what you want your app to do
• Provide context: Include relevant files, URLs, or examples
• Start simple: Begin with basic functionality and add complexity gradually
• Use examples: Show Agent similar apps or features you want to emulate
• Choose your mode thoughtfully: Consider your timeline, certainty about requirements, and development style

​

Managing costs effectively

• Start with free planning: Review Agent’s implementation plan before approving work
• Use simple requests: Break complex projects into smaller, focused tasks
• Set spending limits: Configure usage alerts and budgets to control costs
• Monitor usage: Track checkpoint costs to understand pricing patterns

​

Next steps

Ready to start building with Agent? Here’s how to begin:

1. Create your first app: Use the Create with AI quickstart
2. Choose your build mode: Start with a design for visual-first development or build the full app for comprehensive functionality
3. Learn effective prompting: Follow our vibe coding guide
4. Set up billing controls: Configure spending management
5. Explore Agent 3 features: Try App Testing for self-validation, Max Autonomy for extended builds, and Agents & Automations for building intelligent systems

Learn more about Agent pricing or explore other Replit AI tools.