Install AnythingLLM & chat with your first PDFs
Get a private ChatGPT answering from your own papers in under 20 minutes
AnythingLLM (by Mintplex Labs) is a free, open-source (MIT) all-in-one private AI workspace: drop in PDFs, Word docs, CSVs, or a whole folder and chat with them — with citations — in a familiar ChatGPT-style UI. It is provider-agnostic, so you can run a fully local model (built-in, or via Ollama / LM Studio) or bring your own cloud key (OpenAI, Anthropic, Azure, AWS, Gemini). When you point it at a local model, nothing leaves your machine; your documents are stored and processed locally by default. The fastest way to feel what it does is the one-click desktop app with a built-in local model and a couple of your own PDFs.
- 1 Download the desktop app at anythingllm.com — there is a one-click installer for Mac, Windows, and Linux. Run it; no account is required.
- 2 On first run, pick an LLM provider. Choose the built-in local provider (AnythingLLM downloads a small model for you) so nothing leaves your machine. If you already use Ollama or LM Studio, you can point at those instead — but the built-in option is the zero-setup path.
- 3 Create a workspace. Click 'New Workspace' in the left sidebar and name it something like
lab-papers. A workspace containerises a set of documents into their own chat thread so its context stays clean. - 4 Upload a few PDFs into the workspace — for example two or three papers from your field. AnythingLLM reads and indexes them locally.
- 5 Ask a question grounded in them, e.g.:
Summarise the main finding of each uploaded paper in one sentence, for a biologist new to the topic.Read the reply — if it arrives, your private document-chat is working and nothing was sent to the cloud.
You created a workspace, uploaded PDFs, and got a coherent answer drawn from them — entirely on your own machine, with no account and no cloud calls.
Build a one-workspace knowledge base for a real topic you are reading
A workspace is only as useful as the documents in it. The goal is one clean, focused workspace over a real reading list you can actually interrogate.
Create a workspace for a topic you are genuinely reading about, load 3–5 real documents into it, and ask three questions you could not answer from any single paper alone.
- 1 Create a new workspace named after your topic (e.g.
crispr-off-targets). - 2 Upload 3–5 real PDFs, notes, or a folder relevant to that topic.
- 3 Ask a cross-document question — e.g.
Where do these papers disagree on method, and which gives the most detail? - 4 Ask two follow-ups that build on the first answer, and note whether the assistant stays grounded in your documents.
A named workspace with your real documents loaded and three answered questions that draw on more than one source.