Skip to main content

What’s new

Agent

Intelligent model routing

Intelligent model routing automatically selects the model and effort level that best fit your task, balancing quality and cost. In Power or Max, select Auto to let Replit route each part of your work as its complexity changes. You can also select a specific model and effort level when you want more control.
Agent settings showing Auto, available models, and the Effort control

Select Auto or choose a model and effort level in Agent settings.

Learn more about Auto Mode.

Find your first customers with Growth Skills

Run Growth Skills with Agent to find your app’s first customers. Available partner integrations include Apollo, Clay, SideShift, RevenueCat, Stripe, and PostHog.
Growth Skills catalog showing Apollo Cold Email Launch and Pixel and Web Intent motions alongside a SideShift Consumer and Viral Potential Assessment motion

Browse Growth Skills and run a motion against your app.

Intelligent model routing for Enterprise

Intelligent Model Routing is now the default Agent mode for Enterprise Workspaces. With Auto mode turned on, Replit chooses the best mode and an allowed model for each task. Account admins can control model availability for each Workspace, and you can turn off Auto mode to select an enabled model. The Enterprise Agent mode menu now uses Power and Max and no longer includes Lite.
Agent settings with Auto mode turned on and the model locked to Auto

Turn on Auto mode to let Replit pick the best mode and model for each task.

Learn more about Enterprise model controls.

Integrations

Customize connectors

Enterprise organizations can now use custom OAuth configurations for connectors. For example, configure Gmail with your own Client ID, Client Secret, and scopes instead of Replit’s defaults.
Gmail connector settings with Use custom configurations selected and setup instructions for Google Cloud credentials

Select Use custom configurations for the Gmail connector.

Learn how to configure a custom OAuth app for a connector.