The 25-minute baseline that proves your local OpenClaw works before you build anything real. Install, configure, message, respond — end to end.
End-to-end recipes — concrete things to build with OpenClaw. Every recipe assumes §5.1 hello-world works (your baseline). They get progressively more ambitious from there: a Slack bot for a team, a WhatsApp FAQ answerer for a small business, a Pi-hosted always-on home agent, and a RAG agent over your personal docs. Each recipe declares its verification state — right now they're all sourced-only because Sush is just starting. Expect them to flip to tested-by-sush as he runs them.
5.1 sourced 2026-05-07
5.2 sourced 2026-05-07
A Slack bot that reads your team's standup channel, writes a daily summary, and answers questions about what was discussed. Memory-driven; not a toy.
5.3 sourced 2026-05-07
An always-on agent that handles inbound WhatsApp FAQs for a small business. Pairs with paired-DM-policy + memory for repeat customers.
5.4 sourced 2026-05-07
A Raspberry Pi 5 sitting on your desk, running OpenClaw 24/7, doing daily check-ins and small workflow tasks. The cheapest credible 'I have an agent' setup.
5.5 sourced 2026-05-07
An agent that answers questions about your own documents — meeting notes, journal entries, project docs. QMD memory + a local or hosted model.