A single Markdown file containing the full text of all your sources, with source headers, metadata, and structure preserved. Drop it into any LLM.

Why not just paste my PDFs in?

Pasting drops headings, mangles tables, breaks code blocks, and burns tokens on noise. FileDigest preserves structure and tags sources so the model can reason across them.

Is my data trained on?

No. Files are processed in ephemeral sandboxes and deleted after digest expiry (default 7 days). You control retention. Bring-your-own S3 on Business.

Yes — on Pro and Business. Docling handles scanned PDFs with high-fidelity layout reconstruction.

FileDigestSign in Create a digest

Compile your files into AI-ready context.

Upload PDFs, DOCX files, or folders. FileDigest turns them into one clean, structured, token-aware Markdown digest for Claude, ChatGPT, Gemini, Cursor, and RAG workflows.

Create a digest See example output

Works with Claude ChatGPT Gemini Cursor RAG pipelines

Three values

One tool. Three jobs done right.

Combine

Many files and formats become one coherent Markdown output. Source boundaries preserved, hierarchy intact, no copy-paste.

file_1.pdf

file_2.pdf

file_3.pdf

file_4.docx

digest.md

Clarify

See file names, hierarchy, token counts, parsing status, and metadata. Know exactly what your model is going to read.

attention.pdf11k

BERT.pdf14k

GPT-3.pdf38k

survey-supp.pdf41k

Improve quality

Docling preserves headings, tables, layout, code blocks, and document structure better than any manual paste-in.

Headings312/318

Tables34/34

Code blocks11/11

Layoutpreserved

Structure preservation built on Docling — IBM Research's open-source document AI toolkit.

The workbench

Upload, configure, ship.

Drag in a folder. Inspect every file's tokens and structure. Generate one digest you can drop into any model.

filedigest.app/new

Drop PDFs, DOCX, or a ZIP folder

attention.pdf11.4kparsed

BERT.pdf14.3kparsed

GPT-3.pdf38.2kparsing

Chinchilla.pdf22.1kqueued

No training on your data 30-day default retention GDPR-ready · EU data residency

The output

Markdown your model will actually understand.

Source-tagged sections, clean headings, real tables, intact code blocks. Token report bundled.

# FileDigest Output

> Generated: 2026-04-27 14:32 UTC

> Files: 8 · Tokens: 144,460

## Source 1 — attention.pdf

### 1. Introduction

Recurrent neural networks, long short-term memory and gated recurrent neural networks have been firmly established as state-of-the-art approaches in sequence modeling and transduction problems…

### Table 2 — BLEU scores

| Model | EN-DE | EN-FR |

|-------|------:|------:|

| Transformer (big) | 28.40 | 41.80 |

Report

Tokens144,460

Sources8

Sections312

Tables34

Quality96 / 100

Headings312 / 318

Tables34 / 34

Code blocks11 / 11

Layoutpreserved

Open in dashboard

2.4M

documents compiled

312B

tokens shipped to LLMs

98.7%

structure preserved

< 90s

median compile time

Questions

The shortest FAQ in SaaS.

Free 3M tokens / mo · Pro $15/mo · Business $39/mo · API on request

Stop pasting. Start compiling.

Free for 3M tokens a month. No credit card.

Create your first digest