Built secure from minute one.

Agent HQ runs on a security stack of small specialized layers. Each one watches a specific surface so failures, leaks, and bad input never reach you.

User-facing security layers.

These layers surface in the dashboard so you can see your agent is healthy.

Health indicator

Gateway-keeper + claw-medic feed a green dot on the Home tab. Tap for status.

Schedule status

Claw-jobs and claw-cron show every automated task in the Schedule tab: name, frequency, last run, result, next run.

Connection status

Claw-channel-watch puts a status dot on every connection card so you know what's online.

System updates

Claw-drift logs platform updates in Settings under System Updates.

Audit Log (Pro & Business)

Claw-audit and claw-monitor produce the full security audit and diagnostics view.

Security score (Starter)

Starter and Starter+ see a simplified security score on Home.

Twelve security features. Every one enforced. Every tier.

Every Agent HQ instance ships with all twelve security features turned on. No tier sells you "security as an upsell." If you have an account, you have the stack.

1. Sandbox

Claw-sandbox isolates every code execution in a subprocess your agent cannot escape.

2. Permissions

Claw-perms enforces file permission bits before any read or write your agent attempts.

3. Allowlist

Claw-allowlist gates every outbound HTTP request through a forward proxy. Unknown hosts are blocked.

4. Vault

Claw-vault stores credentials encrypted with rotation. Secrets never appear in chat or logs.

5. Leak detection

Claw-leak scans every outbound response for accidentally-leaked secrets and redacts before delivery.

6. Throttle

Claw-throttle stops runaway loops and rate-limit spirals before they cost you anything.

7. Resources

Claw-resources caps memory, CPU, and process count per container. No agent can starve the host.

8. SSRF protection

Claw-ssrf blocks cloud metadata endpoints and private network probes at the proxy boundary.

9. Injection block

Claw-injection scans inbound payloads for shell-injection patterns and rejects them before execution.

10. TLS enforcement

Claw-tls rejects plain-HTTP requests to unknown hosts. HTTPS-only by default for outbound traffic.

11. Encryption at rest

Claw-encrypt protects credentials and customer data on disk with platform-managed keys.

12. Forensics

Claw-forensics writes a structured audit log of every privileged action: who, what, when, from where.

Plus the layers that keep agents healthy.

These run alongside OpenClaw Shield 12. They handle the boring side of reliability so your agent stays up, your skills stay safe, and dead processes never pile up.

Reaper & session repair

Claw-reaper and claw-session-repair clean up dead processes and recover crashed sessions automatically.

Pin & skills lint

Claw-pin locks dependencies. Claw-skills-lint validates every first-party skill before it loads.

Drift detection

Claw-drift watches for unexpected changes to configuration and platform state, and surfaces them in Settings.

Channel watch

Claw-channel-watch tracks every connected app's health so the dashboard always shows the truth.