Files are read locally. The folder is never uploaded.
This tool runs locally in your browser.
Results are estimates based on your input only; entered values and result contents are not stored by BoringToolsKit.
Your inputs and results stay in this browser. This page may send privacy-bounded aggregate interaction events described in Privacy; it does not send your entered values or result contents.
Use this result
Share the current inputs or ask ChatGPT to explain the calculation in context.
More options
What does this tool show?
Folder to prompt: turn a codebase folder into one organized prompt with full file tree and source code for LLM context. Free, runs locally, no upload.
How this calculator works
Frequently asked questions
How does folder to prompt work?
You upload a folder of files directly in your browser. The tool reads the contents and generates structured prompts based on the file data. Everything is processed locally, no files are sent to a server and no account is required.
Do I need to create an account or sign up?
No. Folder to prompt is a no-signup, browser-based tool. You can use it immediately without registering, and your files stay private since processing happens entirely on your device.
What types of files can folder to prompt handle?
Folder to prompt supports common text and code files such as .txt, .md, .json, .csv, .html, .css, and .js. You drop your folder in, and it extracts content to build ready-to-use prompts for AI tools.
Cite this tool
BoringToolsKit. “Folder to Prompt.” boringtoolskit.com/folder-to-prompt/. Free to reference in articles, syllabi, and answer posts with a link.
Privacy: Inputs and results stay in this browser. Any future sponsored recommendation or advertisement will be clearly labeled and kept separate from the calculation.
What the tool reads, skips, and counts
Files are chosen by extension first. The default filter lists 11 patterns covering JS, TS, TSX, JSX, Python, Markdown, JSON, YAML, HTML and CSS, so a folder heavy with images or binaries contributes nothing before a single file is opened.
A blocklist then removes paths that leak secrets or burn budget: .git, node_modules, .env, private key files, and any name containing credentials, secret, token, .pem, .p12 or .key. Binary content is caught by a null byte and skipped instead of pasting garbage into the prompt.
Each surviving file is capped at 200 KB by default, adjustable from 1 KB up to 2000 KB, and the summary line reports files read, total characters, and an estimated token count.
Token math and when to split a folder
The estimate divides total characters by four. Eight files averaging 32 KB give 262,144 characters, which is about 65,536 tokens, and that already sits near the working context of a 128K model once the system prompt and your own instructions are added.
Raising the size cap to 2000 KB does not make a large folder fit, it only lets bigger single files through. Walking a whole source tree usually crosses the context limit anyway, so split by subfolder or narrow the extension list instead.
The file tree at the top costs characters but saves a round trip, because the model can name a path instead of guessing one. Keep it on for small folders and switch it off once the tree alone runs past a few thousand characters.