Hermes Agent: local model setup and desktop features

This walkthrough covers Hermes Agent's desktop features and the speaker's approach to choosing models for different tasks. Bot mode supports separate profiles with their own memories, personalities and tools. The speaker recommends organizing bots by model rather than job title and says bot mode is available only in Hermes desktop.

For local LLM setup, the tutorial follows Settings, Providers, then "run models locally." According to the speaker, Hermes checks the computer's hardware, recommends a model and can download, load and test it before creating a bot. A 16 GB Mac Mini is one example discussed, though the speaker cautions that smaller computers suit simpler tasks.

The desktop tour also covers an always-on-top HUD with screen access, voice interaction that requires an OpenAI API key, and the "Hey Hermes" wake phrase. Subagent windows expose activity and allow changes to models and skills. The speaker uses recurring stock research to illustrate cron memory, saying it helps avoid repeating earlier research.

The closing comparison explains personal preferences: Hermes for managing computers and tasks involving several models, ChatGPT Work for technical work, and Grockbot for general knowledge tasks on a phone. The speaker identifies Hermes' lack of a mobile app as a limitation and mentions using its iMessage integration.

The video includes a HubSpot sponsorship.