OpenClawd.ai

The site claims 'Complete data ownership, everything stays on your machine' and the privacy policy states 'your data remains on your device and is not transmitted to our servers unless you explicitly configure external integrations.' However, the privacy policy also reveals they collect IP addresses, use Google Analytics and Google Tag Manager, and may use data for 'personalized content and advertisements.' No mention of encryption at rest, tenant isolation, or log sanitization. The self-hosting claim is undermined by the fact that the claimed GitHub repository does not exist, so users cannot verify what the software actually does with their data.

No mention whatsoever of prompt injection protection, sandboxing, memory integrity, or human-in-the-loop controls. The site describes 'Native System Access' where agents can 'run command-line instructions, and execute personalized scripts' with the ability to 'Adjust security settings from isolated environments to full system privileges.' This suggests the agent can escalate to full system access, but no safeguards against hijacking are documented. The absence of any security documentation and the non-existent GitHub repo make it impossible to verify any protections.

Users are instructed to configure API keys for Claude, ChatGPT, and Gemini during onboarding via 'openclawd onboard to establish API credentials and service connections.' No information is provided about how these credentials are stored -- no mention of encryption, secrets management, vault systems, or credential rotation. The ToS places full responsibility on users for 'maintaining the security of your systems, credentials, and data.' The GitHub repo where one could audit credential handling does not exist.

The site describes the agent as 'Action-oriented (executes tasks autonomously)' with capabilities including sending messages, browsing the web, filling forms, and running system commands. No mention of rate limiting, spending caps, kill switches, approval workflows, or behavioral monitoring. The feature description mentions 'Power users can enable the AI to create and implement its own enhancements,' which suggests self-modifying agent behavior with no documented guardrails.

The ToS explicitly states the service is 'AS IS' and 'AS AVAILABLE' without warranties. Users are told they are responsible for 'Creating and maintaining backups of your data and configurations.' No backup tools, export functionality, or data portability features are documented. The claimed documentation site (docs.openclawd.ai) and support site (support.openclawd.ai) both fail to resolve, and the GitHub repository does not exist, raising serious concerns about project continuity.

The pricing model is relatively transparent: self-hosted with API costs of '$5-50 based on usage patterns.' The site clearly lists hardware options ($0-$599) and API cost ranges. However, there is no mention of spending caps, usage alerts, or tools to monitor API consumption. The ToS states 'We reserve the right to modify these Terms at any time.' The score is slightly higher here because self-hosting inherently limits provider-side billing surprises.

The privacy policy lists a contact email (support@openclawd.ai) but the support subdomain does not resolve. No physical address, company registration, or team information is provided. The About page redirects to the main page with no company details. The site states 'OpenClawd AI is an independent project and is not affiliated with OpenClaw or OpenClaw AI' but gives no information about who operates it. No incident response process, no breach notification timeline, no audit logging, no GDPR compliance details beyond generic rights listing.

The site recommends installing via 'curl -fsSL https://openclawd.ai/install.sh | bash' -- a classic insecure pattern that pipes a remote script directly into bash without verification. The install.sh URL does not even serve a valid script (it returns the main page HTML). The GitHub repository at github.com/openclawd/openclawd returns a 404. No dependency scanning, no SBOM, no tool vetting process, no signed releases or checksums are mentioned. Users have no way to audit what they would be installing.

No security page exists on the site. No mention of MFA, TLS configuration, access controls, penetration testing, or security certifications. The site uses HTTPS but provides no information about platform hardening. Since it is positioned as self-hosted software, the platform security depends on user infrastructure, but the provider offers zero guidance on secure deployment. The non-existent documentation site means no security best practices are available.

The site makes numerous claims that cannot be verified: it claims to be open-source but the GitHub repo does not exist; it links to documentation that does not resolve; it provides an install command that does not work; it references 'over 100 applications' of integrations but no integration docs exist. Testimonials from 'Sarah Chen,' 'Marcus Rodriguez,' etc. appear fabricated (stock photo-style names with generic quotes). The FAQ contradicts itself -- some answers describe OpenClawd as the software itself, while others say it is 'a social layer where AI agents can share updates.' This pattern of unverifiable and self-contradictory claims severely undermines trust.