Replit Cloud Agents let non-engineers submit tested pull requests to company codebases

Replit Cloud Agents for existing repositories are now available to all Pro and Enterprise customers, letting people outside engineering propose changes to a company's core codebase as tested pull requests. Replit announced the release on its blog and describes it as part of what it calls the Self-Driving Company.

How it works

A user describes the change they want in plain language. The Cloud Agent finds the relevant code and makes the change, while a live preview of the app updates next to the chat. The user can request adjustments before anything goes to engineering.

When the change looks right, the agent opens a pull request in GitHub or a merge request in GitLab that follows the team's skills and contribution guidelines. If a human or agent reviewer asks for changes, the agent makes them and updates the request.

Testing in isolated sandboxes

Before a pull request reaches a reviewer, Cloud Agents check the change in isolated cloud sandboxes. According to Replit, the agent opens the running app in a browser and clicks through the change as a customer would. If something fails, it fixes the code and tests again. It then attaches screenshots and video of the final version to the pull request.

Review feedback becomes skills (early access)

Replit's self-improving workspace memory turns review comments into guidance for agents. When feedback is worth keeping, the agent proposes an update to the team's skills (the written instructions agents follow in a codebase) or drafts a new one. Engineers approve each update before it takes effect. This feature is in early access.

Replit says it has used workspace memory internally to reduce token usage by 51% and to cut the time agents take to complete a common data task from 16 minutes to 1 minute, without reducing output quality. These are Replit's own figures.

Shared environments

Replit now offers shared environments. An admin sets one up, and a Cloud Agent explores the codebase and configures an initial virtual machine that runs the app. The environment includes the team's repos, runtimes, skills, tools and dependencies, and updates automatically from the main branch.

Replit says environments support dozens of programming languages, microservice architectures with several repos running together, Docker images, databases, queues and caches. Credentials come from external secrets managers through a credential broker. Contributors do not see the setup.

Slack and mobile entry points

Users can mention or DM @Replit in Slack to investigate an issue or turn a proposal into a pull request, follow progress in the thread, and continue in Replit when needed. The result is the same tested pull request with video in GitHub or GitLab. Microsoft Teams support is listed as coming soon. Pull requests can also be started from the Replit mobile app.

Replit also says anyone in a company can now get answers grounded in the actual codebase, such as a support lead investigating why a customer keeps hitting an error. It states that its agent on the company codebase was rolled out internally at the start of March 2026, and that access to code is one of the main drivers of increased work done on Replit in 2026.

Availability

• Cloud Agents for existing repositories: all Pro and Enterprise customers. Repositories connect at replit.com/code-repos.
• Slack integration: all customers. Setup is in Workspace Settings, where users link a Replit account and choose a workspace.
• Workspace memory with skill updates: early access.
• Webinar: Replit is hosting a getting-started session on November 5.

What this means

The release targets two bottlenecks Replit identifies: contributors cannot show that a change works, and engineers lack time to onboard them. Attaching a tested change with video to the pull request moves verification to before review, so reviewers receive something they do not have to run themselves to understand.

The control points stay with engineering. Engineers approve every skill update before it takes effect, and Replit states that engineering still decides what ships. The shared-environment design means one admin-configured setup serves all contributors, which addresses the duplicated setup work Replit says leads people to build on outdated code. The efficiency figures for workspace memory come from Replit's internal use and have not been independently reported.