Cursor’s Expanded Self-Hosted Cloud Agent Infrastructure
Cursor has taken a significant leap in the realm of cloud technology by expanding its self-hosted cloud agent infrastructure. This new update introduces dynamically scheduled worker pools, allowing development teams to manage the machines that execute tool calls, while still starting and managing agent sessions through Cursor. Since the general availability of self-hosted cloud agents on March 25, 2026, the company has provided a blend of control and flexibility for organizations navigating the complexities of cloud development.
Enhanced Control for Development Teams
With the updated infrastructure, development teams have more control over their environments. Cursor’s architecture allows the agent loop, inference, and planning to continue being managed through its cloud, while file edits, terminal commands, repository operations, and other crucial tasks can now be executed on machines that the customer manages. This means teams can tailor their computing environments to best fit their specific needs while benefiting from Cursor’s intelligent orchestration.
Seamless Connection Through Outbound HTTPS
One of the core features of this architecture is the long-lived outbound HTTPS connection that allows workers running on customers’ infrastructure to connect with Cursor. This secure connection ensures that command results are sent back to Cursor for subsequent inference without initiating inbound connections to the customer’s network. This adds an extra layer of security, ensuring that critical operations remain contained within the organization’s private environment.
Use Cases for Self-Hosted Deployments
Cursor anticipates various scenarios where teams would benefit from self-hosted machines rather than relying solely on managed environments. Access to network-restricted resources, specialized hardware, and custom operating systems are just a few examples where self-hosted setups shine. Furthermore, existing build pipelines can be leveraged without significantly disrupting development workflows. This flexibility opens up pathways for teams to harness internal resources that may not be accessible through public domains.
CI/CD-Style Worker Pools
Among the standout features introduced in this update are the dynamically scheduled team pools. These allow workers to serve requests across an entire team or enterprise context, rather than being confined to a single machine or repository. Requests are placed in a named pool and are claimed by available workers, promoting efficient resource utilization.
Organizations can customize pools for various execution environments, such as GPU machines for intensive tasks or Macs tailored for iOS development. Additionally, Cursor’s deployment options for Kubernetes and Google Cloud Run enhance the ability to scale as needed based on real-time demand. This type of flexible scheduling model is reminiscent of traditional CI/CD systems like GitLab Runner and GitHub Actions, but Cursor adds its unique layer by routing agent workloads instead.
Idle Worker Management
Cursor has also introduced the capability to hibernate idle workers—an efficient resource management technique. Organizations can snapshot and stop an idle machine, and the environment can be restored if a follow-up request is made within a specified reconnect time. This feature is essential for optimizing both costs and resource availability.
Security: A Top Priority
While Cursor offers innovative solutions for self-hosted deployments, the company is adamant about keeping security at the forefront. The orchestration, inference, and planning remain under Cursor’s control, while customer-operated workers manage terminal commands, code changes, and requests to internal services.
Information does flow between the worker and Cursor, with tool results being returned via the outbound connection and artifacts generated during sessions also sent back. However, Cursor allows administrators to block the upload endpoint for artifacts, a move that can help safeguard sensitive information, even while continuing tool operations.
Infrastructure Agnostic: A Variety of Hosting Options
Cursor’s self-hosted workers can run on various infrastructures, including offerings from AWS Lambda, Cloudflare, and others. Cloudflare’s implementation, for instance, puts each assigned Cursor session in an isolated container, maintaining an environment where commands and file changes execute seamlessly without exposure to security risks.
Agent Flexibility Beyond the Developer’s Desk
Perhaps one of the more innovative aspects of Cursor’s update is its capability to run agent sessions outside of individual developer workstations. With shared pools, agent sessions can be initiated through various platforms like GitHub, Slack, and Linear, allowing for a more integrated approach to development workflows. The update ensures these agents can modify code, run builds, and even automate the preparation of pull requests—all happening asynchronously in a managed infrastructure.
Expanded Operating System Support
In response to the diverse needs of developers, Cursor now supports self-hosted machines operating on both Linux and Mac systems. This dual compatibility is especially beneficial for workloads requiring specific desktop environments or Apple hardware, such as iOS development. Agents can even interact with browsers, take screenshots, and perform other tasks traditionally restricted to individual developer environments.
Conclusion of Features
Cursor continues to offer its managed virtual machines as the default environment for cloud agents, but the expanded self-hosted capabilities provide a vital alternative for organizations needing more control. Whether it’s to secure internal resources, utilize specialized hardware, or maintain unique development environments, much can be gained from this carefully tailored evolution in Cursor’s service offerings.
Join the Future of Development
For those looking to explore more about the tools and frameworks shaping modern development, the AI & Big Data Expo presents an opportunity to connect with industry leaders and innovators. With sessions focused on machine learning, data pipelines, and cutting-edge AI applications, it’s a chance for developers to dive deep into the future of technology.
With the rapid evolution of self-hosted environments, Cursor’s enhancements may well represent a pivotal moment in the way organizations approach cloud development.