EZClaw
LivePay-per-hour OpenClaw hosting with zero API key setup required
Security Score: 5.5/100 — Basic
EZClaw is a consumer-focused managed OpenClaw hosting service priced at $0.05/hr, targeting non-technical users who want a personal AI assistant on Telegram/WhatsApp. The site is well-designed but almost entirely marketing-focused with virtually no security documentation. There is no dedicated security page, no technical documentation, and no details about infrastructure hardening. The privacy policy and terms are standard boilerplate with heavy liability disclaimers. The terms allow service discontinuation, pricing changes, and account termination without notice. The Google Ads tracking tag with all consent defaults set to granted contradicts the privacy policy's claim of no third-party tracking cookies. The architecture claim of 'Your Data, Your Server' is misleading given this is a fully managed Kubernetes deployment — users do not actually control the underlying server. Credit: the prepaid credit pack model naturally limits billing surprises, the privacy policy has specific data retention periods, and breach notification within 72 hours is committed to.
10 risk categories scored 1-10 × evidence weight. Based on our methodology, grounded in OWASP Agentic Security, NIST CSF 2.0, and CIS Controls.
Homepage claims 'Your own isolated cloud server — not a shared service. Your conversations stay private and secure' and the privacy policy states 'We explicitly do NOT collect or have access to: Your AI conversations (they stay on YOUR server).' However, this is marketing language with no technical details about isolation mechanisms. The privacy policy mentions Kubernetes for deployment but provides no specifics about namespace isolation, network policies, or encryption at rest for agent data. Google Analytics tag (AW-17518297760) is present on every page with all consent defaults set to 'granted' including ad_storage and ad_personalization.
No information found anywhere on the site about prompt injection protection, sandboxing for code execution, human-in-the-loop controls, memory integrity, or any agent hijacking mitigations. The site focuses entirely on ease of setup and has zero security-specific content.
Privacy policy states 'API keys and tokens needed to deploy and configure your server' are collected as configuration data, and 'Your API keys (we pass them directly to your server)' and 'SSH keys generated during setup are stored only in your browser session storage.' Also states 'Server configuration — Stored encrypted, retained while account is active.' These are partially reassuring claims but lack technical detail about encryption methods, key management, or whether credentials are visible in logs.
No information found about rate limiting, spending caps, kill switches, permission scoping, behavioral monitoring, or any controls to prevent agents from taking unauthorized actions. The FAQ answer 'What is the difference from just using ChatGPT' mentions 'memory of your conversations' and 'set reminders' but no guardrails.
Terms state 'WE ARE UNDER NO OBLIGATION TO MAINTAIN, SUPPORT, UPDATE, OR CONTINUE OPERATING EZCLAW. THE SERVICE MAY BE DISCONTINUED ENTIRELY AT ANY TIME WITHOUT PRIOR NOTICE.' No mention of backups, data export, or disaster recovery. The only mitigation is the claim that 'Termination of your EzClaw account or discontinuation of our Service does NOT affect your deployed server,' but since agents run on their managed Kubernetes, this claim is contradictory.
Pricing is clearly published at '$0.05/hr per agent' with credit packs ($10/~200h, $25/~500h, $50/~1000h, $100/~2000h). Payment is via Stripe with prepaid credits, which provides a natural spending cap. However, terms state they can 'Change features, pricing, or availability without prior notice.' The pay-as-you-go credit model inherently limits surprise billing since you pre-purchase credits.
Privacy policy includes 72-hour breach notification: 'In the event of a data breach affecting your personal information, we will notify you via email within 72 hours of becoming aware of the breach.' GDPR/CCPA rights mentioned with 30-day response commitment and contact email (privacy@ezclaw.app). However, no audit logging, no incident response details beyond breach notification, and governing law is Delaware with mandatory arbitration and class action waiver.
No information found about dependency scanning, MCP server vetting, SBOM, build pipeline security, or any supply chain security measures. The site links to upstream OpenClaw GitHub and docs but provides no details about how they vet updates or manage dependencies in their managed deployment.
Dashboard login uses email/password only with no mention of MFA, rate limiting, or session management details. No security page exists (returns 404). Privacy policy mentions 'essential cookies for authentication and session management' but no technical details. The claim 'No third-party tracking or advertising cookies are used' contradicts the Google Ads tag (AW-17518297760) found on every page with all consent defaults granted.
No information found about hallucination warnings, approval workflows, output verification, undo capabilities, or any mitigations for AI misinformation. Terms include an AI disclaimer acknowledging 'AI may generate content that is inaccurate, incomplete, misleading, or factually wrong' but this is purely legal CYA with no technical mitigation.
Key Features
- ✓Pay-per-hour pricing (no monthly commitment)
- ✓No API keys required — AI included via OpenRouter
- ✓Claude, GPT, Gemini models via OpenRouter
- ✓Telegram + WhatsApp integration
- ✓Automatic updates and maintenance
- ✓Credit pack system
Integrations
Strengths
- +Lowest barrier to entry — no API keys, no monthly commit
- +Pay-per-hour means you only pay for actual use
- +AI models included (no BYOK friction)
- +Credit packs up to $100 for heavy users
Weaknesses
- −Limited to Telegram + WhatsApp
- −No SSH access
- −Can get expensive for 24/7 usage ($36/mo)
- −New entrant, limited track record
Verdict
Unique pay-per-hour model perfect for occasional users. No API key setup friction. Gets expensive at 24/7 usage ($36/mo) but cheapest for part-time use.