WordPress for AI agents. Curated blueprints that turn OpenClaw into a purpose-built personal agent — without a PhD in DevOps.
The big 4 AI companies are building personal agents that know everything about you — emails, calendar, tasks, health, finances. They store it on their servers. They train on it. You have zero sovereignty.
OpenClaw is the escape hatch — the most powerful open-source agent framework, running in a Docker container you control. But it's nearly impossible to operate: ~13,500 tokens of config across 11+ files, 14 silent landmines, and most deployments abandoned within a month.
Today you choose between surveillance AI (polished, easy, you own nothing) or raw framework (sovereign, powerful, months of expertise). Nobody makes the sovereign option usable.
ClawHQ is an agent platform for OpenClaw. It forges purpose-built agents from blueprints — complete operational designs that configure every dimension of the agent for a specific job. Choose a blueprint, customize it, and ClawHQ forges a hardened, running agent. You get a Signal, Telegram, or Discord UI. We do the rest.
Everything in OpenClaw is either a file or an API call. ClawHQ controls all of it programmatically — identity, tools, skills, cron, integrations, security, autonomy, memory.
┌─────────────────────────────────────────────────────────────┐
│ Cloud │
│ Managed hosting · Remote monitoring · Blueprint library │
│ ─── optional — the product works without this ─── │
├─────────────────────────────────────────────────────────────┤
│ Blueprints (the product) │
│ Complete agent designs that configure EVERYTHING │
│ in OpenClaw for a specific job. │
├─────────────────────────────────────────────────────────────┤
│ Platform (table stakes) │
│ Install · Harden · Launch · Ops │
│ Same for every agent — acquire engine, secure it, │
│ keep it alive, back it up, update it safely │
└─────────────────────────────────────────────────────────────┘
npm install -g clawhq # Install
clawhq init --guided # Pick a blueprint, connect services
clawhq up # Deploy your agentSecurity hardening is automatic — no opt-in required. See docs/QUICKSTART.md for the full walkthrough.
A blueprint configures identity, tools, skills, cron, integrations, security, autonomy, memory, and egress for a specific job. One example:
# configs/blueprints/email-manager.yaml
name: Email Manager
use_case: "Run my emails"
tools: [email, calendar, tasks]
skills: [email-digest, morning-brief, schedule-guard]
cron: inbox check every 15 min, daily digest at 8am"Run my emails" becomes inbox triage, morning digests, auto-reply with approval gates, and calendar protection — fully configured, hardened, and running.
See docs/CONFIGURATION.md for the full blueprint schema and all available options.
Active development. 59 CLI commands, ~53,000 lines of TypeScript, 52 test files. The platform layer (install, harden, deploy, doctor, backup, update, skills), blueprint customization, the installer, and a web dashboard (clawhq dashboard) are implemented. Cloud provisioning and remote deploy updates are in progress.
Core bet: people will choose a sovereign AI agent over a big-tech one — if the sovereign option isn't dramatically harder to use.
| Document | Description |
|---|---|
| Architecture | Three layers, six modules, zero-trust remote admin, skill system |
| Quickstart | From zero to a working agent in under 10 minutes |
| Configuration | Blueprint schema, skill schema, every config option |
| Problems | Why OpenClaw is hard and what ClawHQ fixes |
| Roadmap | What's built, what's next, honest assessment |
| Contributing | How to contribute blueprints, skills, and code |
| Changelog | Development history and notable changes |
| Product | Product design: problem, personas, user stories |
| OpenClaw Reference | Engine internals, 14 landmines, integration surfaces |
Simon Plant — Building AI infrastructure tools. Open to roles in AI platform engineering and developer tools.
Licensed under the Apache License 2.0.