AI
Coder and SpaceX's Cursor Bridge AI Coding to Regulated Industries
A new partnership lets enterprises deploy advanced coding agents on their own infrastructure, solving the security friction that has locked developers out of powerful tools.
The New Stack
The integration problem#
Coder and SpaceXAI have partnered to launch Coder Agent Relay, a service that addresses a critical gap: organizations in heavily regulated sectors—banking, life sciences, defense, aerospace, government—want to use advanced AI coding tools like Cursor, but cannot send their source code and execution to external cloud services.
The solution splits the workload. Cursor handles inference and planning in the cloud, while actual tool execution runs on the customer's own infrastructure through Coder's self-hosted environments. Code, secrets, and internal services never leave the organization's network, enabling strict compliance without abandoning developer experience.
Why adoption has stalled#
Coder CEO Rob Whiteley describes a stark inequality within his own company: just 1% of engineers account for 40% of token spending. The real issue isn't cost—it's uneven adoption. Experienced developers are leveraging AI agents effectively while others lag behind, creating what he calls "a widening gap between the (agentic) haves and have-nots."
More damaging, regulated enterprises simply block Cursor entirely when security teams review it. "The worst-case is the common case: blocking Cursor outright," Whiteley explains. Not because Cursor is insecure, but because cloud-hosted agents conflict with organizational architecture requirements. Teams that once celebrated productivity gains from Cursor suddenly lose access, embodying what developers sarcastically call "this is why we can't have nice things."
How the partnership solves it#
The technical approach is straightforward but powerful. Each Coder workspace can host a Cursor agent that maintains an outbound connection to Cursor's cloud infrastructure. Platform teams provision and manage these workspaces alongside existing developer environments, scaling regulated AI tooling across the organization.
Agent environments are sandboxed and ephemeral, scoped to single tasks. Security policies block suspicious behavior at the environment layer before the model even needs to refuse a request. Every action is logged—what the agent accessed, executed, changed, and attempted but was blocked from—enabling audit trails without manual reconstruction.
Filling a market gap#
Whiteley argues that the real blocker hasn't been capability, but deployment architecture. Cursor has supported self-hosted agents since March, and many customers were already attempting custom integrations themselves. Formalizing this partnership addresses "the customer should never be the systems integrator" principle, giving enterprises Cursor's interface and experience without surrendering control.

