Shared responsibility model

Security in MultiClaw follows a shared responsibility model. MultiClaw secures the cloud infrastructure and the desktop app; you secure your local environment, credentials, and workspace configuration. Understanding each party's role helps you protect your workspace effectively.

MultiClaw's responsibilities

MultiClaw secures the cloud infrastructure that powers MultiClaw Cloud. This includes the physical security of data centres via AWS, server and database hardening, and network-level protections. MultiClaw Cloud stores data on AWS infrastructure, which encrypts underlying storage volumes at rest. Sensitive values including agent configuration and credentials are additionally encrypted at the application layer using AES-256-GCM before being written to the database.

All connections between the gateway and MultiClaw Cloud are encrypted with TLS. Desktop app updates are signed with minisign and verified before installation. MultiClaw patches and maintains the desktop app and MultiClaw Cloud on an ongoing basis.

MultiClaw Cloud enforces access controls and maintains an audit trail of key platform events. If a security incident is confirmed to have affected your data, MultiClaw will notify the relevant supervisory authority within 72 hours as required by GDPR Art. 33. MultiClaw will also notify you directly without undue delay if the breach poses a high risk to your rights (Art. 34).

Your responsibilities

Keep the desktop app updated

Auto-updates are enabled by default. Do not disable them, as updates include security patches. If you need to verify which version you are running, check the desktop app's About screen.

Keep OpenClaw updated

OpenClaw is installed and updated separately from MultiClaw. Keep your OpenClaw installation current by following the update instructions for your platform (for example, brew upgrade openclaw on macOS). MultiClaw does not distribute or update the OpenClaw binary on your behalf.

Protect your LLM API keys

Your LLM API keys are stored in MultiClaw Cloud and encrypted at rest using AES-256-GCM. You control which services receive those keys. If you suspect a key has been compromised, rotate it immediately in your LLM provider's dashboard and update it in MultiClaw Cloud. Only connect services you trust.

Secure the local OpenClaw data directory

The local OpenClaw data directory is protected by your operating system's file-system permissions. Enable full-disk encryption on any machine running the desktop app. Keep the machine's operating system updated and apply security patches promptly.

On macOS, confirm FileVault is on in System Settings → Privacy & Security. On Windows, check Settings → Privacy & security → Device encryption or search for BitLocker. On Linux, verify LUKS encryption is active.

Manage workspace member access

Invite only trusted users and revoke access when members leave. Two roles exist: Owner and Member. Periodically review the member list by navigating to your workspace's Users page (/{your-workspace}/users) in MultiClaw Cloud.

Configure agent guardrails

MultiClaw provides secure defaults, but you are responsible for adjusting approval flows, scope limits, and execution permissions to match your organisation's risk tolerance. Review guardrail settings whenever your security requirements change.

Vet third-party MCP servers

Before adding a third-party MCP server, verify its source and review its permissions. MultiClaw cannot audit or vouch for third-party servers. You are responsible for any access those servers receive.

Shared responsibilities

Incident detection

MultiClaw monitors infrastructure for anomalies. You are responsible for monitoring your own agent activity. If you notice unexpected behaviour — agents running tasks you did not approve, unusual output, or activity at unexpected times — review the audit trail and revoke agent access if necessary.

Configuration

MultiClaw ships secure defaults for all settings. When you change a default (for example, disabling auto-updates or granting an agent broad file access), you accept responsibility for the security implications of that change. Review any configuration change against your organisation's security policies before applying it.

What this model does not cover

This model describes the division of security responsibilities between MultiClaw and you. It does not guarantee that either party's measures will prevent every possible incident. No security programme can eliminate all risk.

If you are unsure where a specific responsibility falls, contact MultiClaw support. For the technical detail behind the protections described here, see Security overview, Data encryption, and How credentials and secrets are stored.