Tool guide
OpenClaw: use cases for small business
OpenClaw is a **self-hosted** AI assistant gateway: you run it on your own machine or VPS, message it from chat apps you already use, and keep sessions and workspace data local. It is **not** a hosted chatbot product with a monthly OpenClaw plan. Facts stick to official OpenClaw docs fetched on 16 Oct 2026. For the Hermes head-to-head, see Hermes Agent vs OpenClaw. For rent-vs-run, see Self-hosted vs hosted AI agents.
What to hand to an agent first
What OpenClaw is
MIT open source from the OpenClaw Foundation. A Gateway bridges chat channels to an agent. No paid OpenClaw SaaS tier — you pay compute and model keys.
Morning brief first
Schedule one Automation that summarises calendar and unfinished items to Telegram or Slack. Read-and-report until the brief is useful.
Inbox triage drafts
Sort mail into reply / FYI / ignore and draft replies. Nothing sends until you approve. Prompt injection in email is a real risk.
Lead follow-ups
Quiet leads get a next-step draft that references the last conversation. You approve every send for the first weeks.
Security defaults
Gateway binds to loopback; unknown DMs use pairing; groups are allowlisted. Run openclaw security audit before exposing beyond localhost.
Cost shape
Software $0 (MIT). You still pay a VPS or always-on machine plus model API (or local models). We do not invent VPS dollar amounts.
Not sure which jobs fit your business?
Build my free AI plan →Ready-made use cases
- 01CEO morning brief in 5 minutes
One message each morning: calendar, urgent emails, money in/out, and the 3 things that move the business today.
~3h/week · estimate - 02Inbox triage with drafted replies
The agent sorts your inbox into reply / FYI / ignore and drafts replies in your voice for you to approve.
~5h/week · estimate - 03Never miss a lead follow-up
The agent checks your CRM or inbox for leads that went quiet and drafts the next follow-up.
~4h/week · estimate - 0424/7 FAQ support in your voice
The agent answers common customer questions from your docs and hands off anything tricky.
~8h/week · estimate - 05Invoice chaser that gets you paid
The agent tracks unpaid invoices and sends polite, escalating reminders.
~2h/week · estimate
Example: Grok Bot setup
Grok Bot routine: CEO morning brief in 5 minutes Schedule: Daily 7:30 Every morning, read my calendar for today, my unread emails from the last 24h and yesterday's revenue. Send me a brief: (1) meetings with one prep note each, (2) emails that need my reply with a suggested one-line answer, (3) money in/out, (4) the 3 highest-leverage tasks for today. Keep it under 200 words. Rules: draft first, never send or post anything external without my approval. Keep summaries short.
Grok Bot, ChatGPT, Claude and n8n versions: all templates
Setup steps
- Decide you will self-host (always-on PC or VPS) — OpenClaw has no hosted SaaS tier.
- Install via the official installer, run onboarding, then install the Gateway as a background service.
- Connect one chat channel (Telegram is often fastest) and one model provider; keep DMs on pairing/allowlist defaults.
- Turn on one scheduled automation (morning brief) for three days before connecting inbox or send tools.
- Run openclaw security audit before exposing the Gateway beyond loopback; read their Security docs for your version.
Keep these human
- Money moves (payments, refunds, ad spend) stay on approval until you trust the routine.
- Anything sent in your name needs your OK at first.
- One gateway is not a hostile multi-tenant security boundary — separate trust domains need separate gateways.
- Don't dual-install OpenClaw and Hermes on day one.
- Don't paste passwords, card numbers or sensitive personal data into prompts.
FAQ
What should an AI agent do first for OpenClaw?
What OpenClaw is: MIT open source from the OpenClaw Foundation. A Gateway bridges chat channels to an agent. No paid OpenClaw SaaS tier — you pay compute and model keys. Morning brief first: Schedule one Automation that summarises calendar and unfinished items to Telegram or Slack. Read-and-report until the brief is useful. Inbox triage drafts: Sort mail into reply / FYI / ignore and draft replies. Nothing sends until you approve. Prompt injection in email is a real risk.
What should stay human?
Money moves (payments, refunds, ad spend) stay on approval until you trust the routine. Anything sent in your name needs your OK at first. One gateway is not a hostile multi-tenant security boundary — separate trust domains need separate gateways. Don't dual-install OpenClaw and Hermes on day one. Don't paste passwords, card numbers or sensitive personal data into prompts.
Is there a ready-made template?
Yes. Start with "CEO morning brief in 5 minutes": copy-ready Grok Bot, ChatGPT, Claude and n8n templates are at https://whatshouldmybotdo.com/use-cases/ceo-morning-brief#templates.