Summary
Matthew Berman presents 21 concrete OpenClaw use cases, sharing the actual prompts used to build each one and demonstrating them live. The video functions as both a capability showcase and a practical build guide—every workflow was created through natural language instructions to OpenClaw itself, with no direct coding required. Integrations covered include Gmail, Google Calendar, YouTube Analytics API, Instagram, X/Twitter, HubSpot, Asana, Fathom (meeting transcripts), Todoist, Slack, GitHub, and Brave Search.
Standout workflows include a custom CRM that ingests contacts from email and calendar; a business advisory council where eight specialized AI agents (financial, marketing, growth, and others) run in parallel each night, analyze 14 business data sources, and deliver ranked recommendations to Telegram; and a security council that continuously audits the OpenClaw instance itself for vulnerabilities and prompt injection risks. The knowledge base workflow ingests articles and X/Twitter threads using a multi-tier API fallback chain and stores everything in a hybrid SQL-plus-vector database for natural language retrieval.
Berman also explains OpenClaw’s memory architecture in detail: daily conversation notes are distilled into a memory.md file, vectorized for RAG search, and fed back into identity.md each session—allowing the assistant to evolve its understanding of the user’s preferences over time. The soul.md and identity.md files enable context-aware personality switching between casual personal interactions and formal business communications, depending on which Slack channel or Telegram topic is active.
📺 Source: Matthew Berman · Published February 17, 2026
🏷️ Format: Hands On Build
