Cursor Lets AI Agents Run on Your Infrastructure

Cursor is bringing its AI-powered cloud agents directly into enterprise infrastructure. The move allows companies to run these autonomous coding assistants entirely within their own networks, a critical step for organizations with strict security, compliance, or complex development environment requirements.

These self-hosted cloud agents maintain the benefits of Cursor's AI tools while ensuring codebases, tool execution, and build artifacts never leave a company's controlled environment. For teams dealing with sensitive data or specialized internal tools, this offers a significant advantage.

The capability is particularly impactful for highly regulated sectors like finance, where data exfiltration is a major concern. It also benefits companies with mature, internally secured development environments that rely on specific caches, dependencies, or network endpoints not accessible externally.

Why Self-Hosted?

Many enterprises operate under stringent regulations preventing sensitive data from leaving their premises. Previously, teams needing such control often invested heavily in building and maintaining their own custom background agents.

Cursor's self-hosted solution aims to alleviate this burden. It allows organizations to retain their existing security postures and build environments while Cursor manages the AI orchestration, model access, and user experience. This frees up engineering resources to focus on development rather than infrastructure maintenance.

Graham Fuller, Senior Software Engineer at Brex, noted the significance: "Now with self-hosted cloud agents, we can give them access to the infrastructure needed to run our test suites and validate changes with our internal tools."

Yokoyama Tatsuo, Deputy Manager of SRE & MEPAR at Money Forward, added, "Given our strict security requirements as a financial services provider, self-hosted support is something we've been eagerly awaiting."

Same Product, Your Infrastructure

Self-hosted cloud agents mirror the functionality of Cursor's hosted offering. This includes isolated remote environments for parallel processing, the ability to launch agents from various platforms like the Cursor editor, Slack, or GitHub, and automation triggers via events or schedules.

Extensibility through plugins, subagents, and rules is also supported. Team permissions allow granular control over agent access and management across an organization. Future updates will enable agents to produce videos and screenshots of their work, with the ability for users to remotely take over agent desktops for direct interaction and edits.

Ben Kraft, Software Engineer at Notion, highlighted the advantage for large codebases: "In large codebases like Notion's, running agent workloads in our own cloud environment allows agents to access more tools more securely and saves our team from needing to maintain multiple stacks." This capability marks a significant stride toward making self-hosted AI agents truly enterprise-ready.

How It Works

The system relies on 'workers', processes that establish outbound HTTPS connections to Cursor's cloud. This architecture eliminates the need for inbound ports, firewall adjustments, or VPNs, simplifying deployment. When an agent session is initiated, Cursor's harness manages AI inference and planning, dispatching tool calls to the worker for execution on the company's hardware.

Results are fed back to Cursor for subsequent inference cycles. Each agent session utilizes a dedicated worker, launched with the simple command agent worker start. These workers can be configured for continuous operation or single-task use.

For large-scale deployments, Cursor provides a Helm chart and Kubernetes operator to manage worker scaling, updates, and lifecycle. A fleet management API supports autoscaling on non-Kubernetes infrastructure. This approach simplifies the integration of secure AI development environments into existing enterprise workflows.

The company also provides insights into advanced topics such as enterprise AI coding assistants, further cementing its position in the evolving AI landscape.