Lima v2.0 introduces plugin infrastructure, GPU acceleration via krunkit, and Model Context Protocol (MCP) tools for sandboxing AI agents. The release expands Lima's focus beyond containers to include secure AI workflows by running agents inside VMs to isolate them from direct host access. New features include progress flags, simplified mount syntax, environment variable propagation, UDP port forwarding, and multi-user support. The plugin system enables third-party extensions for hypervisors, CLI commands, and URL schemes.
Table of contents
What is Lima ?Updates in v2.0GPU accelerationModel Context ProtocolOther improvementsExpanding the focus to hardening AIAI inside LimaAI outside LimaGetting started: AI inside LimaSort: