Vagon review

Vagon is a hosted cloud-computer service: it provisions remote virtual machines with configurable CPU and GPU performance, then streams the desktop to a browser or desktop app. It is not presented as a general-purpose cloud API, a local library, or a self-hosted server. The machine is where users run demanding creative, engineering, and development software, or supported agents such as Claude Code, Codex, and OpenCode.

That puts Vagon below the application layer, as remote compute and desktop delivery rather than an application backend. The homepage lists 140+ preinstalled apps and 11 GPU and CPU performance options, which users can switch between. Teams can also use its application-streaming product to deliver Unreal Engine and Unity experiences to other devices without code. Vagon says its machines are isolated virtual machines and connections are encrypted. For a codebase that needs a programmable compute control plane, teams should first establish whether Vagon supports the required automation and integration surface.

Vagon's homepage describes access through a browser and desktop apps, not a code-first API workflow. Its listed integration surface includes remote desktops, application streaming, and AI agents running next to installed applications. The homepage demonstrates MCP connected to Blender, and features guides for using Claude Code or Codex with apps including Premiere Pro, After Effects, Blender, and SolidWorks.

The vendor names no API style, SDK, programming language, or CLI on the homepage. Developers building provisioning or session management into a product should confirm SDK support in Vagon's docs and ask the vendor which operations can be automated. The homepage also describes Vagon Interactive Streaming Protocol for streamed sessions, with up to 4K and 60 FPS; that is a streaming protocol, not evidence of a general-purpose application API. Teams evaluating integrations should verify authentication, session lifecycle controls, file transfer, and error handling with the vendor before committing.

Vagon lists no free plan or free trial. Its lowest listed monthly price is $3.99 per computer for the Teams Trial Plan, which includes up to 1 hour of usage, 75 GB disk storage, 25 GB file storage, and Planet and Spark performance. Despite the plan name, it is a paid plan. The other monthly options are $30.07 for All Performances, $35.59 for Office Works, $59.59 for Creative Works, and $96.74 for Create & Export, each per computer. Those plans include up to 40 hours, with storage and performance selections varying by plan.

When included hours run out, extra usage is billed by the minute. Listed rates range from $0.25 per hour for Sand to $11.97 per hour for Lava, vary by region, and depend on performance. Outbound transfer costs $1.50 per additional 10 GB after the first 10 GB per month. No annual prices are shown. For workloads with uneven demand, estimate runtime by performance tier and region, plus transfer, before comparing monthly plan prices.

Vagon is offered as a hosted cloud-computer service. The vendor does not publish an open-source license or describe a self-hosted edition, so teams should treat the service as hosted-only unless Vagon confirms otherwise. The homepage says users control their data, but does not describe a data-export format or a migration process.

The main lock-in risk is operational: applications, agent workflows, and project files may be arranged inside Vagon machines, while streaming and performance configuration are tied to its service. Before moving sensitive or valuable work there, ask about bulk export, persistent storage behavior, machine images, and what remains accessible after cancellation. Apps and project assets that are independently licensed and stored in portable formats should be easier to move than workflows that depend on Vagon-specific session or streaming behavior.

Vagon lists 20+ regions worldwide, 1 Gbps+ connection speed inside each machine, and streaming up to 4K at 60 FPS with full-color support. It also says users can switch among 11 GPU and CPU performance options. These are vendor-published capabilities, not a substitute for checking latency from your users' locations or testing the applications and peripherals your workflow requires.

Plan limits matter operationally: monthly plans include either up to 1 hour or up to 40 hours, and extra use is billed by the minute. Ask Vagon about uptime commitments, support response targets, backup and recovery, compliance certifications, and regional data handling before depending on it for production work. One Reddit user reports that a camera was not connected to their virtual desktop, while audio worked, and says they had not received a support response. That is one user's report, but it makes peripheral compatibility and support escalation worth checking.