Create a KVM VM through the REST API, wait for guest readiness, run code via
/exec, and tear it down when done. No direct SSH integration in your controller.
Per-minute billing — destroy to stop charges. Full GUI for humans too.
# 1. Obtain a token (once) POST https://api.litevps.dev/api/auth/login { "email": "agent@example.com", "password": "..." } ← { "token": "eyJ..." } # 2. Create a VM using current plan and template IDs POST https://api.litevps.dev/api/vps Authorization: Bearer eyJ... { "name": "task-runner", "plan_id": "current-plan-id", "template": "current-template-slug" } ← { "vps": { "id": "d4e1f2a3-...", "state": "running", "ready": false } } # 3. Wait until vps.ready is true, then run a command POST https://api.litevps.dev/api/vps/d4e1f2a3-.../exec { "command": "python3 solve.py --input data.json" } ← { "stdout": "result: 42\n", "stderr": "", "exit_code": 0, "duration_ms": 1823 } # 4. Destroy it — billing stops, resources released DELETE https://api.litevps.dev/api/vps/d4e1f2a3-... ← 200 { "message": "VPS deleted" }
Batch jobs, agent sandboxes, internal APIs, background workers — most workloads don't need a dedicated public IPv4. Outbound internet, proxied inbound access, and private networking cover almost every real use case. We pass the savings directly to you. See pricing →
Billed per minute — running or stopped, a VM accrues charges as long as it exists. Destroy it the moment you're done and billing stops immediately. No hourly rounding, no minimum commitment. Your agent's cost matches its actual lifetime exactly.
A small REST workflow creates a VM, reports guest readiness, runs code via /exec, and tears the VM down. Command responses include stdout, stderr, and exit_code.
Every VM uses KVM with its own guest kernel and private network address. Outbound internet is included; public access is provided through managed proxy and port mappings.
Predictable JSON in, JSON out. Every endpoint is documented at /docs with request/response schemas and a full general guide. Designed to be consumed by both humans and language models.
Checkpoint a VM mid-task and restore it later. Useful for agents that need to branch execution or retry from a known-good state without re-provisioning from scratch.