Player not loading? Watch on YouTube
This tutorial walks through setting up OpenClaw in Atomic Chat and choosing a model for its chat interface. The speaker describes installation as a one-click process. In the AI models section, they activate an already downloaded model identified as Gemma 4, with a displayed size of 17 GB, then send a message through the OpenClaw dashboard. They report a reply within a minute and describe this local LLM setup as slower than their usual models.
The guide also covers API connections. OpenRouter lets the speaker select from free and paid models using their own API key. For Ollama, they create a key, save it in Atomic Chat, and test the connection. An earlier cloud request fails before the key is connected; a later request succeeds. Although the speaker initially calls the selected cloud model local, the subsequent demonstration explicitly uses Ollama cloud models.
The speaker's claim of free ongoing use has a qualification: Ollama cloud usage must stay within its token limits. They suggest switching to downloaded models when those limits are reached. Viewers who want to run models locally also get a brief tour of backups, command approval, voice input, and messenger settings. The tutorial ends by showing where to upload a custom skill through the Claw Hub skills section.
