Why Combine Files Before Sending Them to AI?
Modern AI assistants like ChatGPT, Claude, Gemini, and Copilot impose limits on the number of files you can attach in a single prompt — even on paid Pro plans. Worse, when you upload many small files, the model often loses track of which content came from which file. Combining your files into one well-structured text bundle solves both problems: you get past the file-count limit, and the AI gets clear, labeled context it can reason about.
How This AI Context Bundler Works
Drop in any number of text-based files (CSV, JSON, YAML, Markdown, source code, logs, configs, plain text). The tool concatenates them into a single .txt file with a structured wrapper around each source — including the filename, size, line count, and content. You choose the wrapper format that your AI parses best. Everything happens in your browser; no file ever leaves your device.
Choosing the Right Output Format
XML-style tags (<file name="...">) are the default because Anthropic explicitly recommends them for Claude, and GPT-4o / Gemini parse them just as reliably. Markdown headings (## File: name) are great when you also want the bundle to be human-readable in a viewer. Plain delimiters (===== FILE: name =====) use the fewest tokens, useful when you're tight against a context limit. Custom templates let advanced users define their own wrapper using {name}, {content}, {ext}, and {size} placeholders.
Token Estimation Across Models
Different models tokenize text differently, but a reliable approximation is ~4 characters per token for English and ~2 characters per token for code or non-Latin scripts. The tool estimates token usage and shows whether your bundle fits within common context windows: GPT-4o (128K), Claude 3.5 Sonnet (200K), Gemini 1.5 Pro (2M). If you're over the limit, enable "strip blank lines" or "strip comments" to shrink the bundle without losing semantic content.
Common Use Cases
Sharing GSC exports with an AI analyst — bundle Queries.csv, Pages.csv, Countries.csv, Devices.csv into one upload so the AI can cross-reference them. Code reviews — combine related source files (a component, its test, its types) into a single context for a focused review. Log triage — concatenate multiple log shards with timestamps preserved. Documentation Q&A — bundle README, CONTRIBUTING, and config files so the AI can answer onboarding questions accurately.
Privacy Guarantee
Your files are read directly into your browser using the File API. Nothing is uploaded, no server is contacted, and the tool works fully offline once the page has loaded. Even the token-count math runs locally. This makes the bundler safe for sensitive material like internal logs, proprietary code, financial CSVs, and unpublished writing.
Frequently Asked Questions
What file types can I combine? Any text-based format: .txt, .md, .csv, .json, .yaml, .yml, .xml, .html, .css, .js, .ts, .tsx, .jsx, .py, .go, .rs, .java, .rb, .sh, .sql, .log, .env.example, and similar. Binary formats (PDF, DOCX, images) aren't supported — convert them first.
Is there a file-count or size limit? No artificial limit. The practical ceiling is your browser's available memory; bundling thousands of large files may slow the page.
Can I reorder files? Yes — drag the handle on each file row to reorder. AI models read top to bottom, so put the most important context first.
Will the AI understand the wrapper? Yes. All four output formats use unambiguous, well-known delimiter conventions that LLMs reliably parse. The default XML format is the most robust.
Does it strip my comments? Only if you tick the option. By default, your files are concatenated verbatim.