OpenShell sandbox image pre-configured with Ollama for running local LLMs.
- 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_HOSTis set for OpenShell provider discovery - Claude Code — Pre-installed (
claudecommand) - Codex — Pre-installed (
@openai/codexnpm package) - Node.js 22 — Runtime for npm-based tools
- npm global — Configured to install to user directory (works with read-only
/usr)
docker build -t openshell-ollama .openshell sandbox create --from ollamaupdate-ollamaOr auto-update on startup:
openshell sandbox create --from ollama -e OLLAMA_UPDATE=1