Chatd is a desktop application that lets you chat with your own documents using a local large language model — no cloud, no uploads, no accounts. It is open source under the MIT license, has earned over 1,000 stars on GitHub, and bundles everything needed to run completely privately on your own machine.
Most document-chat tools make you wire up a model server yourself before you can ask a single question. Chatd removes that friction by packaging the local LLM runner with the app: download the executable, point it at your files, and start asking questions. Under the hood it uses Ollama to serve models such as Mistral 7B and Llama 2, and if you already run an Ollama instance, Chatd can connect to that instead. Retrieval-augmented generation grounds the answers in your documents, and because inference happens locally, sensitive files — contracts, personal notes, unpublished research — never leave your computer. The project was most recently updated in mid-2024 and remains a tidy, self-contained example of local-first document AI.
Chatd’s design centers on a single promise: your documents stay yours. There is no account system, no analytics pipeline, and no cloud fallback quietly uploading files for processing — the model, the embeddings, and the document index all live on your machine. That makes it usable in situations where cloud AI tools are simply not an option, such as legal or medical offices handling privileged files. It also works fully offline once installed, which suits travel, fieldwork, or any environment where connectivity is unreliable or untrusted.
Chatd is free and open source under the MIT license. There is no paid tier; the only cost is your own hardware running the local model.
Chatd fits privacy-conscious professionals, researchers, and writers who want to interrogate document collections without sending confidential material to a third-party AI service.