After Meta took Muse Code out of beta for larger engineering tasks, Ollama said the harness already works out of the box. `ollama launch muse` wires Muse Code to models on Ollama local and cloud, as an alternative to Meta’s `curl | bash` installer. Teams that want Meta’s coding agent on open weights—without leaving Ollama’s local or zero-retention cloud path—can treat Muse as another first-class launch target.
Key Takeaways
- ✓Muse Code is out of beta; Meta’s installer is `curl -fsSL https://dev.meta.ai/install.sh | bash`.
- ✓Ollama’s path is `ollama launch muse` against local and cloud models.
- ✓Lets Meta’s engineering agent run on Ollama-hosted open weights without a separate inference stack.