Are my files uploaded when I use these tools?

No, the file contents stay in the browser tab. Processing starts the moment you pick a file and runs on your device. For file tools there is no server path that could carry the contents to kittokit.

Why can the AI not read my PDF properly?

Usually the text is an image inside the PDF. A scanned or photographed document holds pixels, not readable characters. Assistants often extract nothing from it and answer anyway.

What do I do when a document is too long for the model?

Measure first, then split. The word counter gives characters, words and an estimated token figure before you paste. The estimate runs close in practice but is not an exact count for any particular model. You place the cuts yourself.

How do I check whether JSON from an AI answer is valid?

Paste it and format it. The formatter shows immediately whether the text parses and where it breaks. The usual causes are truncated answers and code fences that got copied along.

How do I strip location data from a screenshot before pasting it?

Look first, then remove. Phone photos often carry coordinates, a timestamp and the device model. Both steps run locally, and the image itself goes nowhere.

Can I pass a result straight into the next tool?

No, and that gap is deliberate. Each tool downloads its result and you hand the file to the next one. With no server in between, there is nowhere for a file to sit between two tools.

AI Workflow Tools: Prepare Files for AI

Offered by Carina in Höxter, Germany. Report an issue or ask a question.

Where are you stuck?

Into the model12 toolsIt will not read the file at all, or reads a few pages and invents the rest.PDF to Markdown DOCX to Markdown XLSX to Markdown
Out, before it travels4 toolsNames, account numbers or location data are still in there, and you are not allowed to upload that.Markdown Cleaner Remove Metadata EXIF Viewer
Check the answer11 toolsThe JSON will not parse, the regex matches everything or nothing, and two runs disagree.JSON Formatter JSON Diff Text Diff
Use the answer3 toolsThe chat is full of ### and ***, and none of it belongs in the finished document.Markdown to PDF Markdown to DOCX Markdown to HTML

Four runs, end to end

See every tool

The order matters more than any single tool.

Recorded meetingAn hour of audio should become a summary, without the recording going anywhere.
Scanned contractsThe model currently receives blank pages, and the names inside are nobody else’s business.
Broken JSONThe agent returns JSON, your code rejects it, and two runs look different.
CSV exportThe CSV arrives with broken encoding and the wrong delimiters, and it is too long to paste.

Each step downloads its result and you hand it to the next tool. There is no automatic handover.

Or run the model yourselfFor when not even the text should leave the machine: paste a model link, get the start commands and a rough memory figure, plus a checklist you can copy.

Into the model

12 tools

PDF to MarkdownPulls out the text with headings and lists intact. Also from scans that otherwise arrive as a blank page.DOCX to MarkdownA Word document as clean text, with the structure still standing.XLSX to MarkdownSheets as pipe tables, which a model reads far more reliably than a spreadsheet file.CSV to MarkdownDetects the delimiter and returns a readable table.HTML to MarkdownA saved page without the nav and script leftovers that otherwise eat context.Image to TextA screenshot or photo as text you can drop straight into the prompt.Audio TranscriptionTurns a recording into a transcript, with timestamps if you want them.Fast TranscriptionFaster than real time when the recording is English and long.Video to AudioPulls the audio track out of the video so transcription has something to work on.CSV CleanerFixes encoding garbage, wrong delimiters and comma-versus-dot decimals before import.Describe ImageA description of an image, for when the rest of your chain is text only.Word CounterCharacters, words and a token estimate before you paste. Estimated, not exactly counted.

Out, before it travels

4 tools

Markdown CleanerReplaces names and numbers across up to 100 Markdown files. You supply the terms.Remove MetadataTakes camera, timestamp, author and coordinates out of images and PDFs.EXIF ViewerShows you what is in there first. Then you decide whether it has to go.Clean Tracking URLCuts links back to what they link to, without the campaign tail.

Check the answer

11 tools

JSON FormatterTells you whether it parses, and exactly where it breaks.JSON DiffCompares two runs structurally instead of by eye.Text DiffShows what actually changed between two answers.YAML to JSONSurfaces where YAML quietly turned your NO into false.TOML to JSONChecks generated project configs without flattening the datetime types.CSV to JSONTurns a table answer into something your code accepts. Leading zeros survive.JSON to CSVThe way back, when the result has to land in a spreadsheet.Base64Decodes the blob a tool call returned, without pasting it anywhere.JWT DecoderThe call returns 401. Here are the claims and the expiry.SQL FormatterBreaks up the 200-character line before you run it.XML FormatterMakes nesting visible, and broken nesting obvious.

Use the answer

3 tools

Markdown to PDFTurns the answer into a document you can send.Markdown to DOCXHeadings become real Word styles, not just bold text.Markdown to HTMLA ready fragment for the CMS, or a file you can open.

Why local, when a hosted model is the next stop anyway

That is exactly why. What you send the assistant is a decision you make. What you take out first never arrives there, and cannot be retained.

.	In the browser tab	At an upload service
Where the file sits	In the tab’s memory. Opened, processed, closed again.	On someone else’s server, usually stored for a while.
Who can read it	Nobody but you. There is no server path for the file contents.	The provider has to receive the file, or it cannot do anything.
With no connection	Keeps working once the page has loaded.	Nothing happens without a network.
The limit	Very large files depend on your device’s memory and speed.	A server has more of both, and also sees the file.

The honest trade: for very large files a server is faster. For anything confidential, that is the price local processing is worth.

Common questions

Are my files uploaded when I use these tools?: No, the file contents stay in the browser tab. Processing starts the moment you pick a file and runs on your device. For file tools there is no server path that could carry the contents to kittokit.
Why can the AI not read my PDF properly?: Usually the text is an image inside the PDF. A scanned or photographed document holds pixels, not readable characters. Assistants often extract nothing from it and answer anyway.Straight there:PDF to Markdown Image to Text
What do I do when a document is too long for the model?: Measure first, then split. The word counter gives characters, words and an estimated token figure before you paste. The estimate runs close in practice but is not an exact count for any particular model. You place the cuts yourself.Straight there:Word Counter
How do I check whether JSON from an AI answer is valid?: Paste it and format it. The formatter shows immediately whether the text parses and where it breaks. The usual causes are truncated answers and code fences that got copied along.Straight there:JSON Formatter JSON Diff
How do I strip location data from a screenshot before pasting it?: Look first, then remove. Phone photos often carry coordinates, a timestamp and the device model. Both steps run locally, and the image itself goes nowhere.Straight there:EXIF Viewer Remove Metadata
Can I pass a result straight into the next tool?: No, and that gap is deliberate. Each tool downloads its result and you hand the file to the next one. With no server in between, there is nowhere for a file to sit between two tools.