Local Models
Local-model agents run on a Mac mini, a GPU box or a home server, with no per-token bill and no data leaving the building. They suit background loops such as monitoring, triage and nightly reports, where a smaller open model is good enough.
Top use cases
Run a private 24/7 agent on your own server with Hermes and Ollama
Always-on assistant
Hermes Agent points at a local Ollama model. Its built-in scheduler runs recurring jobs and sends results to Slack, Telegram or Discord, with no API keys and no data leaving your machine.
Keep an OpenClaw assistant always on on a Mac mini
Always-on assistant
OpenClaw runs as a background service on a Mac mini. It answers you over WhatsApp, Slack or iMessage, watches inboxes and runs reminders around the clock, and you reach it remotely through Tailscale instead of a public port.
Run a local model as a continuous research loop
Research
An open 27B model runs on your own hardware around the clock, working through a long research task and logging findings, with no cloud cost per iteration.
Schedule watchdog scripts that alert Slack without an LLM
Monitoring
For checks that need no reasoning, such as stock levels, site uptime or failed payments, Hermes runs your script on a schedule and posts the output to Slack, Telegram or SMS, so nothing is spent on a model.
Local Models FAQ
Do I need a powerful computer?
A Mac mini or a PC with a modern GPU runs useful 8B to 30B models. Bigger models need more memory.
Are local models as good as cloud models?
Not for everything. They are strong for routine loops; use a cloud model for hard reasoning.
Why run an agent locally?
Privacy, no per-token cost, and it can run 24/7 on your schedule.