Cursor said the system is designed for organizations with strict security and compliance requirements that cannot allow sensitive data or development workflows to leave their environment. By running agents locally, teams can connect them directly to internal dependencies, caches, and network endpoints, mirroring how engineers or service accounts operate within existing infrastructure.
The company positions self-hosted agents as an extension of its existing cloud agent system, with the same orchestration and model capabilities but with execution handled inside the customer’s environment. This allows organizations to retain their current security models while offloading agent management and coordination to Cursor’s platform.
A worker process connects outbound to Cursor’s cloud to receive instructions, avoiding the need for inbound network changes such as open ports or VPNs. Each agent session is assigned a dedicated worker, which can persist or terminate depending on workload requirements. Cursor also provides tools for scaling deployments, including Kubernetes support and fleet management APIs.
The release comes as enterprises increasingly look for ways to integrate AI agents into development workflows without exposing proprietary data. Customers including Brex, Money Forward, and Notion are already using the system to run coding agents within their own environments, with some deploying workflows that allow engineers to generate pull requests directly from tools like Slack.
Cursor said the update is aimed at making AI coding agents viable for large organizations, particularly those operating under strict regulatory or security constraints, while reducing the need for teams to build and maintain their own agent infrastructure.
This analysis is based on reporting from Cursor.
Image courtesy of Cursor.
This article was generated with AI assistance and reviewed for accuracy and quality.
