xCloud

The FAQ states 'Your OpenClaw instance runs on a dedicated, isolated environment. We use enterprise-grade encryption, automatic SSL certificates, and daily encrypted backups. Your data stays yours – we never access it except for technical support when you request it.' However, no specifics are given about what 'enterprise-grade encryption' means (algorithm, at-rest vs in-transit), and the privacy policy mentions using aggregated traffic data for research ('You consent to xCloud using your traffic data for research purposes'). The privacy policy is generic and WordPress-focused, with no OpenClaw-specific data handling details. No mention of whether agent conversation data is used for AI model training.

The main page explicitly states 'OpenClaw is not a sandboxed agent. It runs on a dedicated VM with full system access, allowing it to install tools, build apps, and operate autonomously.' While running on a dedicated VM provides some isolation between users, the lack of sandboxing within the VM means there are no documented protections against prompt injection, memory poisoning, or code execution exploits. No mention of human-in-the-loop controls, instruction-data separation, or container escape prevention.

Users must bring their own API keys (FAQ: 'you need to bring your own AI provider API key such as OpenAI, Anthropic, Claude Code, OpenAI, or others') and Telegram/WhatsApp bot tokens. There is zero documentation on how these credentials are stored, whether they are encrypted at rest, whether they appear in logs, or how credential lifecycle management works. The features list mentions 'encrypted tokens' and 'auth' in passing ('Automatic security hardening includes encrypted tokens, auth, auto SSL, and strict firewall rules') but provides no technical details about what encryption is used or how tokens are managed.

No information found about rate limiting, spending caps on agent actions, kill switches, behavioral monitoring, least-privilege enforcement, or any guardrails for autonomous agent behavior. The platform explicitly markets full system access and autonomous operation as features rather than risks to mitigate.

The pricing page lists 'Daily automatic backups' as an included feature, and the FAQ mentions 'daily encrypted backups.' The managed hosting page mentions 'Full Server Backup: Automatically schedule & manage backup of entire server and ensure data recovery for any possible disaster.' However, there is no documentation on backup testing, restore procedures, data export capability, or what happens to user data if xCloud shuts down. The T&C states 'Your account and xCloud servers will be removed after 5 days if you do not clear unpaid bills' which is concerning for data retention. The company is xCloud Hosting LLC based in Milton, DE.

Pricing is relatively transparent: plans start at $24/month for Standard (4GB RAM, 80GB SSD) up to $399/month for Business Max (64GB RAM, 1280GB SSD). The T&C provides 30 days notice for service/pricing changes ('Changes or discontinuation of the services may occur with 30 days written notice'). Users bring their own AI API keys so API costs are separate and controlled by the user. However, there are no documented hard spending caps on the xCloud side, and no usage monitoring or alerts are mentioned.

The privacy policy claims GDPR compliance support but with a significant disclaimer: 'While using xCloud, you need to double-check if GDPR compliance is applicable to you or not. If you fail to comply with any GDPR law, xCloud is not liable or responsible for this in any way.' The docs show server and site event logging that tracks activity with serial numbers, event names, initiator names, and status. However, there is no documented incident response process, no breach notification timeline, and no agent-specific action audit trail. Data is stored under US jurisdiction (Delaware LLC).

No information about dependency scanning, MCP server vetting, build pipeline integrity, component inventory (SBOM), or how xCloud manages the OpenClaw software supply chain. The platform is powered by WPDeveloper in partnership with WPManageNinja, but there is no transparency about how OpenClaw updates are validated before deployment, or what third-party components are included.

The login page shows only email/password authentication with no MFA/2FA option visible. The features mention 'auto SSL' and 'strict firewall rules' and the platform has a 7G WAF on Nginx that protects against 'hacking attempts, SQL injections, XSS attacks.' Password requirements exist (minimum 8 characters, uppercase, number, special character). However, the Site Security PRO product (Patchstack-powered) is WordPress-specific and costs extra. No mention of independent security testing, penetration testing, or security audits. The heading 'Enterprise-grade security' on the managed hosting page is unsupported by any specifics.

No information found about approval workflows, output verification, undo/rollback capabilities, prompt injection monitoring, or transparency about AI uncertainty. The platform positions itself as a deployment tool rather than addressing the trust and verification challenges inherent in autonomous AI agent operation.