OpenClaw Launch

Provides some specific technical details: 'Each bot runs in an isolated Docker container with dedicated storage — fully sandboxed from other users' and 'Configurations are encrypted at rest with AES-256-GCM encryption.' Privacy policy states 'We do not use your configurations, conversations, or any instance data to train AI models.' However, no mention of employee access controls, audit trails, or log sanitization. Data stored on Vultr with PostgreSQL but no details on database encryption or access controls beyond 'secure server access controls.'

The FAQ mentions 'isolated Docker container' which implies some sandboxing, but there is no documentation of prompt injection defenses, human-in-the-loop controls, memory integrity protection, output sanitization, or container escape prevention. The claim of isolation is vague regarding agent hijacking specifically — Docker containers alone do not address prompt injection or memory poisoning risks.

FAQ states 'Gateway tokens are auto-generated using cryptographically secure random UUIDs' and configurations are 'encrypted at rest with AES-256-GCM encryption.' The privacy policy says 'We never access your conversations, API keys.' However, there is no documentation of credential leak detection in outputs, credential lifecycle management, rotation policies, least-privilege access models, or whether credentials are excluded from AI model context windows.

No information found on any guardrails against agent misbehavior. No mention of rate limiting, spending caps on agent actions, kill switches, behavioral monitoring, tool classification, or least-privilege/least-agency principles. The platform appears to deploy standard OpenClaw instances with no additional controls documented.

Terms explicitly state 'Instance data is not backed up. You are responsible for exporting any data you need' and 'Docker instances are provided on a best-effort basis with no uptime guarantee.' There is no data export capability documented, no circuit breakers, no provider stability signals (no funding information, no team page). The terms allow modification, suspension, or discontinuation 'at any time.' Score of 1 reflects documented anti-patterns rather than no information.

Pricing is clearly displayed: Lite at \/mo (\ first month) and Pro at \/mo. Models page shows exact per-token API costs with 'No markup — you pay the exact same rates as the original providers.' AI credits are pay-as-you-go with included amounts. However, no hard spending caps, no usage alerts, and terms reserve the right to change pricing without defined notification period — 'We reserve the right to modify these terms at any time.'

No incident response process documented. No breach notification timeline. No audit logging mentioned. No agent action audit trail. Privacy policy mentions data stored on Vultr but no jurisdiction details beyond that. No GDPR compliance statement despite collecting EU user data via OAuth. Only accountability mechanism is a contact email (support@openclawlaunch.com). The claim 'We never access your conversations' is unverifiable without audit controls.

No information found on dependency scanning, MCP server or skill vetting (despite referencing '5,700+ community skills on ClawHub'), build pipeline security, SBOM, or AI provider data policies. The news page reports a critical RCE vulnerability (CVE-2026-25253, CVSS 8.8) in upstream OpenClaw but there is no documentation of how quickly OpenClaw Launch applies such patches or their update process.

Login supports Google OAuth, GitHub OAuth, and email/password authentication. SSL/HTTPS is documented as 'automatic certificate management.' However, no MFA option is visible on the login page. No mention of injection attack prevention, SSRF protection, security testing, or platform hardening. The use of OAuth providers (Google, GitHub) delegates some authentication security but the email/password fallback has no documented protections (rate limiting, brute force prevention).

No information found on approval workflows, output verification, prompt injection monitoring, undo/rollback capabilities, or transparency about AI uncertainty. The platform appears to be a direct deployment of OpenClaw with no additional trust or verification layers added by the hosting provider.