Running AI models on your own hardware just got more useful. LM Studio has shipped a reworked plugin system that adds web research capabilities to locally-run language models, with everything staying on your machine.
Plug-and-Play, No API Keys
The big addition is a web research plugin that lets local models search the web and incorporate results into their responses. This is notable because web access has been one of the main reasons people still reach for cloud-hosted tools like ChatGPT or Perplexity instead of running models locally. LM Studio's implementation runs the entire pipeline on-device, meaning your queries never hit a third-party server.
The plugin architecture itself got a complete rework. Previous versions required manual configuration. The new system is designed as plug-and-play: install a plugin, and it works with whatever model you already have loaded. That's a meaningful improvement for the local AI crowd, where setup friction is still one of the biggest barriers to adoption.
Who This Is For
LM Studio has carved out a strong position as the easiest way to run open-source models like Llama, Mistral, and Qwen on a Mac or PC. Adding a plugin system that extends what those models can do, without sending data to the cloud, directly addresses the two main complaints from local AI users: limited capabilities compared to cloud tools, and the technical skill required to add features.
The web research plugin is the headline feature, but the reworked architecture matters more long-term. A solid plugin system means the community can build extensions for specific workflows (code execution, file analysis, database queries) without waiting for LM Studio's team to ship each one individually. For anyone running local models for privacy or cost reasons, this closes a real gap.