If you’re searching for “local LLM” or “run LLM locally”, you usually want three things:
Privacy (your prompts and data stay on your machine)
Predictable cost (no per-token API fees)
A model you can call from code, automation, and other tools
Zo Computer is a good fit because it’s a real Linux server you control, with an always-available filesystem and the ability to run long-lived services.
This tutorial shows how to install and run a local LLM with Ollama on Zo, then expose it as a persistent service you can use from scripts and Agents.
Prerequisites
A Zo Computer
Enough RAM for the model you want to run
7B–8B models are a reasonable starting point
Bigger models need substantially more RAM
Basic comfort using the terminal inside Zo
Step 1: Install Ollama
Ollama provides a one-line installer for Linux.
Run:
curl -fsSL https://ollama.com/install.sh | sh
Confirm it’s installed:
ollama --version
Step 2: Download (pull) a model
Pick a model from the Ollama library and pull it.
For example:
ollama pull llama3.2
List what you have installed:
ollama list
Step 3: Chat with your model (quick test)
ollama run llama3.2
If you get a response, the core setup is working.
Step 4: Run Ollama as a persistent service on Zo
If you only run ollama run ... interactively, it stops when you close your session. On Zo, the typical pattern is: run the server as a managed service so it auto-restarts.
Start the Ollama server (foreground test):
ollama serve
By default, Ollama listens on 127.0.0.1:11434.
Stop it (Ctrl+C), then register it as a Zo service so it stays up.
From Zo, create a service that runs:
ollama serve
(Zo services are managed from the Services page, and they’ll restart automatically if the process crashes.)^4
Step 5: Call your local LLM over HTTP
Once ollama serve is running, you can call it from inside Zo.
A quick curl test:
curl http://127.0.0.1:11434/api/generate \
-d '{"model":"llama3.2","prompt":"Write a one-sentence summary of Zo Computer."}'
This is the key unlock: you now have a local model you can use from scripts, tools, and Agents.
Step 6: Use it from an Agent (practical pattern)
A simple, high-value workflow is:
An Agent runs on a schedule
It reads files (notes, logs, CRM exports, whatever you keep on Zo)
It calls your local Ollama model to summarize / classify / draft
It writes a result file or emails you the output
If you haven’t used Agents before, start here:^5
Common issues
It’s slow
Use a smaller model.
Reduce how much text you send per request.
Consider running on a Zo machine size with more CPU/RAM.
I want to access Ollama from outside Zo
Start by deciding what you actually need:
If you want a public endpoint, use Zo’s Services system and expose the port intentionally.
If you only want remote access for yourself, SSH port forwarding is often the simplest approach.
(As a default, keep Ollama bound to localhost and only open it up when you have a clear use case.)^3
Suggested next steps
Treat your Zo filesystem as your “model memory”: build a folder of prompts, context snippets, and reusable instructions.
Combine your local model with Zo’s built-in tools (web browsing, file operations, integrations) for hybrid workflows.