Cursor adds cloud agent dev environments for end-to-end tasks
Written byMango
Drafted with AI; edited and reviewed by a human.
![]()
TL;DR
- Cursor now offers configurable development environments for cloud agents, enabling end-to-end task completion.
- Support for multi-repo environments and improved Dockerfile configuration with build secrets are key features.
- Enhanced security includes version history, audit logs, and scoped secrets for each environment.
- Builds hitting the cache now run an impressive 70% faster due to upgraded layer caching.
Cursor has unveiled a significant enhancement for its cloud agents, introducing customizable development environments that bring the familiar setup of a local machine to AI-driven workflows. This update empowers teams to configure complete development stacks for their agents, including cloned repositories, installed dependencies, and necessary credentials, enabling agents to tackle complex engineering tasks from inception to completion within fully controlled environments. This move facilitates the operation of parallelized agent fleets capable of executing tasks end-to-end.
A major advancement is the introduction of multi-repo environments, building on previous multi-root workspace capabilities. This allows developers to define a single environment encompassing all repositories an agent requires for its operations, with the ability to reuse these configurations across different work sessions. This streamlines workflows by ensuring agents have access to all necessary codebases without manual setup for each task.
Further bolstering the system is an improved approach to environment configuration as code. By enhancing Dockerfile-based setups, Cursor now supports build secrets, which provide secure access to private package registries directly within Dockerfiles. These secrets are isolated to the build step and are not exposed to the running agent's environment, adding a crucial layer of security. Additionally, upgraded layer caching means that when a Dockerfile is modified, only the affected layers are rebuilt, leading to builds that hit the cache running up to 70% faster.
Cursor also simplifies the setup process through improved agent-led environment configuration. The platform now actively guides users by asking relevant questions, flagging any missing credentials, and validating the setup to ensure correctness. Agents will always operate within a clearly indicated version of their environment, and in case of configuration failures, they will gracefully default to a base image, ensuring continued operation rather than immediate disruption.
Security and governance have also been significantly enhanced. Each development environment now features its own version history, allowing users to review and roll back changes. Administrators can also restrict rollback permissions solely to themselves. An audit log meticulously tracks all actions performed on environments by team members, providing security teams with comprehensive visibility into all modifications. Furthermore, egress and secrets can be precisely scoped at the development environment level, meaning secrets configured for one environment are entirely inaccessible from any other, reinforcing isolation and security. For detailed information on these advancements, you can explore the Cloud Agent Development Environments Announcement and the Cloud Agent Setup Documentation.
Summary
- Cursor's latest update introduces configurable development environments for cloud agents, enabling end-to-end task completion within controlled setups.
- New features include multi-repo support, improved Dockerfile configuration with build secrets, and significantly faster builds due to upgraded layer caching (up to 70% faster).
- Enhanced governance and security are provided through environment version history, audit logs, and secrets scoped to individual environments.
- Users can find more details and setup guides at Cursor Changelog and related documentation.
Read next

Claude's New Best Practices Boost Browser Automation Accuracy
Anthropic releases best practices for using Claude models in computer and browser automation, focusing on image scaling for improved click accuracy.
Continue reading