Ollama Sandbox

March 23, 2026 · View on GitHub

OpenShell sandbox image pre-configured with Ollama for running local LLMs.

What's Included

  • Ollama — Ollama runs cloud and local models and connects them to tools like Claude Code, Codex, OpenCode, and more.
  • Auto-start — Ollama server starts automatically when the sandbox starts
  • Pre-configured — OLLAMA_HOST is set for OpenShell provider discovery
  • Claude Code — Pre-installed (claude command)
  • Codex — Pre-installed (@openai/codex npm package)
  • Node.js 22 — Runtime for npm-based tools
  • npm global — Configured to install to user directory (works with read-only /usr)

Build

docker build -t openshell-ollama .

Usage

Create a sandbox

openshell sandbox create --from ollama

Update Ollama inside the sandbox

update-ollama

Or auto-update on startup:

openshell sandbox create --from ollama -e OLLAMA_UPDATE=1