Cut Your AI Token Usage by Up to 60%
Your AI can't understand your files.
You're paying for it anyway.
TokenFlow converts PDFs, Word docs, Excel sheets, audio files, and web pages into clean, structured markdown so your LLM can read every document accurately without burning through your token budget or context window.
I Built TokenFlow Because I Saw AI Wasting Time Instead of Saving It
I watched firms feed countless PDFs, PowerPoints, and webpages into LLMs or AI agents for analysis and review.
Brand Strategy Guides. Investor decks. Product pages. The documents that actually matter.
And more often than not, they ended up paying 2-3x more for tokens than they should have and the outputs were still 80% useless because the inputs were broken before they ever got read.
The ingestion layer was the bottleneck. Not the model. Not the prompt.
TokenFlow cleans the document first, so your AI reads structure it can actually use.
Three Steps to AI-Ready Documents:
Step 1: Upload
Drop files or paste a URL. Batch upload supported.
Step 2: Convert
Engine extracts text, preserves structure, outputs clean markdown.
Step 3: Save
Share with ChatGPT, Claude, Gemini, or your AI agent.
What You Get:
PDF to Markdown — Clean, structured output that LLMs understand the first time.
Excel Multi-Sheet — Preview sheet names; pick which to convert.
Audio Transcription — Turn voice recordings into plain text.
URL to Text — Fetch pages stripped of ads, nav, and anything else that isn’t valuable context.
Batch Conversion — Upload a ZIP archive to optimize multiplefiles at once.
Agent Skill — One click and your OpenClaw or Hermes agents automatically convert any file.
Token Counter — See what you saved on each conversion.
Lifetime Stats — Watch your savings accumulate over time.
Supported Formats — PDF, DOCX, PPTX, XLSX, CSV, JSON, XML, HTML, MP3, WAV, OGG, FLAC, M4A, ZIP, URLs, and more coming soon.
Start Free. Scale When You Need To.
Billing by Stripe. Upgrade/cancel anytime.
Questions?
-
Free tools have file size limits, watermarks, and no API access. TokenFlow gives you structured output (markdown or HTML), direct API integration, and configurable security. Batch processing available via the /convert/batch endpoint. You own the pipeline, not the vendor.
-
Yes. Every plan includes an API key for programmatic document conversion. Integrate PDF-to-text extraction, Word-to-markdown conversion, or audio transcription directly into your app, workflow, or data pipeline. See the API documentation for Python, Node.js, and cURL examples.
-
PDF, Word (DOCX), PowerPoint (PPTX), Excel (XLSX/XLS), CSV, JSON, XML, HTML, ZIP archives, and URLs. Audio files: MP3, WAV, OGG, FLAC, M4A. We extract text, tables, and structure — not just raw content. See the full supported formats list.
-
Yes. Files are uploaded over TLS, processed on isolated Fly.io VMs, and converted outputs are stored temporarily in Supabase Storage with tier-based retention. Raw input files are deleted immediately after conversion. We never share documents with 3rd parties. For teams requiring zero retention, contact us about enterprise plans.
-
Raw input files are deleted immediately after conversion. Converted outputs are stored temporarily in Supabase Storage with tier-based retention so you can download results. After retention expires, outputs are permanently deleted. For custom retention requirements, contact us.
-
Use the TokenFlow API. Send your PDF via POST request with your API key, and get structured text back in seconds. Supports multi-page documents, table extraction, and embedded images converted to base64 markdown.
-
Yes. Upload DOCX files to TokenFLow and get clean markdown output with headers, lists, and tables preserved. Ideal for documentation workflows, static site generators, and LLM context preparation.
-
Yes. Upload MP3, WAV, OGG, FLAC, or M4A files and receive full text transcripts. Supports multiple languages. Audio conversion requires a Pro plan or higher.
-
TokenFlow extracts every file from the ZIP and converts each supported document individually. Each successful conversion counts against your monthly quota. If you hit your quota mid-archive, remaining files are skipped with a "quota exceeded" note. Unsupported files (like executables or nested ZIPs) are skipped with an error. The results appear as expandable items in the web interface.
-
Yes. Upload a ZIP archive and each supported file is converted individually. Each successful conversion counts against your monthly quota. Unsupported files are skipped with an error note. If you hit your quota mid-archive, remaining files are skipped. Multi-sheet Excel files inside ZIPs convert all sheets automatically — for sheet selection, convert XLSX files individually via the /convert endpoint with the sheets parameter. For high-volume separate-file processing, use the /convert/batch endpoint or contact us for workflow integrations.
-
Most documents convert in under 5 seconds. A 100-page PDF typically takes 10–15 seconds. Audio transcription runs at approximately 0.5x real-time (a 10-minute file processes in ~5 minutes). All plans share a rate limit of 1,000 requests per hour.