Operate any combination of infrastructure with AI

• Manage resources. Create and configure Global Virtual Clouds (GVCs), workloads, identities, domains, secrets, and policies.
• Deploy applications. Build images and deploy across every region of your virtual cloud.
• Query infrastructure. Metrics, logs, traces, events, deployments, quotas, and the audit trail, in real time.
• Automate workflows. Troubleshoot a failing workload, wire a secret to the service that needs it, configure a custom domain, migrate a Kubernetes manifest, or export a GVC™ to Terraform.

Every agent action passes through the same identity, policy, and audit controls as a human operator.

• Per-org consent. Each AI tool is authorized for the smallest set of orgs its job needs.
• Toolset profiles. Each connection gets one of four fixed profiles (readonly, core, mk8s, or full). Use readonly to let an agent triage a production incident with no ability to change anything.
• Least privilege, by policy. Agents are principals, bound by the same policies as users and service accounts.
• Full audit. Every action, by a human or an agent, is logged and attributable.

Isolated development environments inside your virtual cloud where an agent can clone a repo, install dependencies, run the app, and reach the private services it depends on, with no credentials in its context.

• Ready in 60 to 90 seconds. Choose an image, a size, and an app port.
• Browser IDE and terminal included, plus a shareable URL for the running app. Or open it in VS Code, Cursor, or SSH with cpln sandbox connect.
• Any toolchain. Start from a template, compose an image from runtimes and packages, or bring your own OCI image.
• Scoped access. Attach an identity and secrets at creation; the sandbox reaches private services through Universal Cloud Identity and Cloud Wormhole. The agent never sees a key.
• Suspend and resume. The home directory is on a persistent volume. Suspend scales to zero and bills storage only; resume picks up where you left off.