BriefGPT is a locally hosted, open-source tool that connects your documents to large language models for summarization and question answering, wrapped in a simple graphical interface. Because everything runs on your own machine, you stay in control of your documents and your API keys instead of uploading files to a third-party service. The project is hosted on GitHub, where it has earned roughly 795 stars.
BriefGPT follows the now-familiar “chat with your documents” pattern, implemented with privacy as the headline feature. You point it at your files, it embeds and indexes them, and you can then ask questions or request summaries through a Streamlit interface. It can work in two modes: calling OpenAI’s models with your own key, or running fully locally, using Instructor embeddings together with LlamaCpp or GPT4All models in GGML format placed in a local models folder. That fully local path means even the LLM never leaves your hardware, which suits sensitive documents. The repository documents the environment variables and setup needed for both modes.
BriefGPT is free and open source under the MIT license. If you choose the OpenAI mode you pay only your own API usage; the fully local mode costs nothing beyond your hardware.
BriefGPT suits individuals and small teams who want document summarization and Q&A with LLM help but cannot — or do not want to — upload their files to a hosted AI service.