BunnyQuery is an AI data assistant for your own files. You upload your files as they are, your Bunny reads each one into records, and you ask questions in plain language.
What your Bunny can do
- Store and Read any files. Spreadsheets, Word documents, slides, PDFs, scans, source code and photographs, in any language.
- Answer from all of it at once. For a total, a count or a "list everything", your Bunny reads every matching record and gives you the answer.
- Give the data back. Export hundreds of thousands of records as CSV, TSV or JSONL, built on our servers without a row passing through the AI. The CSV and TSV files carry a UTF-8 marker, so non-Latin text opens correctly in Excel. Or ask for the answer as a written document.
- Share it, or plug it into other tools. Add collaborators, embed the chat on your own site, or reach your data from Claude and any other MCP-compatible AI client.
Why you need it
Your files hold answers you cannot get to. The information is spread across formats that do not talk to each other: photos, PDFs, spreadsheets, documents. To answer one question you would have to open each file, read it, and copy what matters into one place by hand. It takes hours and answers one question.
So you do not ask. BunnyQuery removes that cost: you upload the files as they are, and your Bunny does the reading.
Example: one question, three kinds of file
Your travel receipts from last year are on paper, the invoices arrived as PDFs, and the card statement is a spreadsheet.
You photograph the receipts, download the PDFs and drag it all in. Nothing gets renamed or sorted. Then you ask: How much did I spend on travel last year? You get a number and a breakdown drawn from all three sources, and you can ask where any part of it came from.
Your data is secure from 3rd party AI platforms
Your files are stored in your own project, not on the AI platform. ChatGPT and Claude never receive a copy of your library.
What they get is the piece a request needs, and nothing more. Your Bunny reads each file once, a window at a time, to write down what is in it. After that, answering a question fetches only the records that match it. Storing a thousand files does not mean sending a thousand files to the AI.
Your project runs on your own API key, so your data goes through the platform's API and not through their consumer chat products. Both OpenAI and Anthropic state that data sent through their APIs is not used to train their models. Nothing you upload becomes part of a model.
- Exports never touch the AI. Ask for a CSV and the query runs on our servers, with the rows going straight into the file. Not one of them passes through the model.
- Permissions are enforced by the database, not by the AI. Private records are masked before any result exists, so nobody can talk a model into showing them what they are not allowed to see.
- File links are short lived. Where the AI is handed a link at all, it is minted for that one request and expires in minutes.
Why we believe this is the future
Folders, file names and forms with fixed boxes exist because computers could not read documents. A machine that cannot read a page can still match a file name, and it can add up a column somebody defined in advance. So somebody always had to translate reality into a shape the machine could handle.
That constraint is gone. A computer can now read a page, in any language, and it reads the four thousandth file as carefully as the first.
The organizing still has to happen. A total needs your numbers to still be numbers and your dates to still be dates. Your Bunny does that work as it reads. Data still has to end up organized. What ends is you doing the organizing.
So go and find your messy folder to upload. Enjoy your Bunny!
Baksa, Creator of BunnyQuery