PDFMacroPDFMacro
Tool · Chat with PDF

Chat with a PDF — on your device, not on a server

Ask a document questions, summarise it, or compare what two clauses say. The AI runs locally, so the file you are asking about never leaves this tab.

Processed in your browser

  • The model runs on your machine

    Either the built-in model, which downloads once to your browser, or your own Ollama or LM Studio server if you have one. Every hosted chat-with-PDF service has your document on their server the moment you ask the first question. This one does not.

  • Nothing is uploaded and nothing is indexed

    Text is extracted in the tab and the relevant passages are handed to the model as you ask. No embedding on a server, no copy retained, no account. After the first model download it works with the network disconnected.

  • Search is free, and often the better tool

    If you know roughly what was written, full-text search finds the passage instantly with no model and no plan. Chat earns its place when the answer is a judgement rather than a passage — a summary, a comparison, what a clause actually means.

Chat with PDF is part of PDFMacro Pro. You can open the tool and see how it works on any plan; asking a question is what requires the licence. Search inside a PDF is free and needs no model.

Common questions

Frequently asked questions

Where does the AI run when I chat with a PDF?
On your machine. Either the built-in model, which downloads once to your browser and then works offline, or your own Ollama or LM Studio server if you connect one. There is no request to us and none to a third-party AI provider, which is the whole difference between this and the hosted chat-with-PDF services — with those, the document is on someone else's server the moment you ask your first question.
Is my document sent anywhere to be indexed?
No. The text is extracted in the tab and the relevant passages are given to the model as context at the moment you ask. Nothing is uploaded, nothing is embedded on a server, and no copy is retained after you close the tab. The one deliberate exception is a local AI server you have connected yourself, where excerpts go to that server — your machine — and nowhere else.
How is this different from Search inside PDF?
Search finds passages; chat answers questions about them. Search is free, instant, needs no model and is the right tool when you know roughly what was written. Chat can summarise, compare clauses and explain what something means, at the cost of a model download and a paid plan. Many people use Search first and only reach for chat when the answer is not a passage but a judgement.
Does it work on a scanned document?
Only once the pages carry a text layer. A scan is an image until OCR has run over it, and there is nothing for the model to read until then. Make Searchable does that on-device too, so a scanned contract can go through the whole path — OCR, then chat — without leaving your machine.
Does Chat with PDF upload my PDF to a server?
No. The file is opened, processed and saved back by code running in this browser tab — there is no upload endpoint on our servers to send it to. You can watch the Network panel in your browser's developer tools while you work and confirm nothing carrying the document leaves the page.
Is Chat with PDF free?
Chat with PDF is part of PDFMacro Pro. You can read what it does and open the tool on any plan; running it requires a subscription. The tool opens and explains itself on any plan; asking a question is what requires the licence.
Does it work offline?
Yes. Once the page has loaded, every tool keeps working with the network disconnected, because the processing never depended on a server round-trip. AI features that use an on-device model need one download the first time; after that they run offline too.
Related
Search Inside PDFOCROutline & Bookmarks

Chat with PDF runs in your browser. Try it now.

No account, no upload, no watermark on the way out. Open a file and see for yourself — the network panel is right there if you want to check.

Open the workspace See pricing