What it was

OpenClaw became the fastest-growing open-source agent framework while being nearly impossible to operate correctly: thousands of tokens of configuration across 11+ files, dangerous defaults, memory bloat, and a wave of exposed instances and CVEs. Hosting providers sold convenience — default-config agents on a VPS. Nobody solved sovereignty.

ClawHQ compiled blueprints — complete operational designs covering identity, tools, skills, cron, integrations, security, autonomy, memory, and egress — into hardened, running agents on your own hardware. Security hardening was automatic, not opt-in. Underneath sat a platform layer that was the same for every agent: install, harden, launch, ops.

Why it’s retired

ClawHQ was a great experiment, and it did its job: it taught me what operating sovereign agents actually requires. But I was its only serious user, and the generalized platform was more machinery than one operator needs. I now manage my production agent directly, with the hardening patterns applied by hand where they matter.

What carried forward

Nearly everything. Sterling — my advisory trading agent — runs on the exact posture ClawHQ was built to produce: read-only containers, credential isolation, DNS allowlisting, egress firewalling, config rendered from version-controlled sources. The OpenClaw architecture reference on this site is the distilled operator knowledge.

The repository stays up at its final working snapshot.

Repository →