This video is about the "Step 2" of using OpenClaw. That is: just getting it to actually do something useful. In this video, I explore how the TOOLS.md and skills work... and how various models can and cannot handle OpenClaw. We use the Remotive API and a custom node.js CLI tool, as well as ask Boardy AI on Whatsapp to help us make introductions (agents talking to agents!). We talk a bit about OpenClaw architecture and how to get the most out of it without burning through all of your cash using a Claude Opus or Sonnet model. This is the beginning of a series about leveraging OpenClaw to do and manage all kinds of tasks. Additionally, we talk about connecting OpenClaw to Gmail, Trello, Twitter and Hashnode securely (Don't use Clawhub!)
Models used in video: together/meta-llama/Llama-3.3-70B-Instruct-Turbo
and anthropic/claude-sonnet-4-5-20250929
Blog it and I am writing (from the video based on what we do each day): opencloser.hashnode.dev