Initial release: Nutrient PDF to Markdown CLI

Fast, accurate Markdown from PDFs — locally, with no cleanup required.
Built for Claude, Codex, Pi, Cursor, Gemini CLI, RAG pipelines,
and document-heavy automation.

- 0.007s per page (90x faster than docling, 37x faster than pymupdf4llm)
- 0.92 reading order accuracy (best in class)
- 0.88 overall extraction accuracy
- Free for up to 1,000 documents/month

Includes CLI wrapper, installer, benchmark data (200 docs),
and documentation.

Author: jdrhyne

.github/workflows/validate.yml

@@ -0,0 +1,27 @@
+name: Validate
+
+on:
+  push:
+  pull_request:
+
+jobs:
+  validate:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 20
+
+      - name: Check shell syntax
+        run: npm run check
+
+      - name: Verify installer copies local wrapper
+        run: |
+          tmpdir="$(mktemp -d)"
+          INSTALL_DIR="$tmpdir/bin" ./install.sh
+          test -x "$tmpdir/bin/pdf-to-markdown"
+
+      - name: Verify npm pack surface
+        run: npm pack --dry-run

.gitignore

@@ -0,0 +1,5 @@
+node_modules/
+*.tgz
+.DS_Store
+sample-report.pdf
+report.md

LICENSE.md

@@ -0,0 +1,79 @@
+# Nutrient Free Use License
+
+Copyright (c) Nutrient.io
+
+## 1. License Grant
+
+Nutrient grants you a limited, non-exclusive, non-transferable, revocable license to use the accompanying software (the "Software") for internal business or personal purposes, subject to the terms of this License.
+
+## 2. Free Usage Tier
+
+You may use the Software free of charge to process up to 1,000 documents per calendar month.
+
+For the purposes of this License:
+
+A "document" means any individual file submitted for processing by the Software, regardless of format.
+
+Each processing event counts as one document, including repeated or duplicate processing of the same file.
+
+## 3. Commercial Use
+
+You must obtain a commercial license from Nutrient if you:
+
+Process more than 1,000 documents per calendar month
+
+To obtain a commercial license, contact: sales@nutrient.io
+
+## 4. Permitted Uses Within Free Tier
+
+Subject to the 1,000 documents per calendar month limit, you may:
+
+Use the Software for internal or external purposes
+
+Incorporate the Software into applications, including hosted, SaaS, OEM, embedded, or white-labeled solutions
+
+Provide the Software's functionality to third parties
+
+## 5. Restrictions
+
+You may not:
+
+Circumvent, disable, or interfere with usage limits or licensing controls
+
+Use the Software to provide a service or product that competes with Nutrient's commercial offerings
+
+Remove or alter any copyright, trademark, or proprietary notices
+
+Use the Software in violation of applicable laws or regulations
+
+## 6. Ownership
+
+The Software is licensed, not sold. Nutrient and its licensors retain all right, title, and interest in and to the Software.
+
+## 7. Termination
+
+This License automatically terminates if you fail to comply with any of its terms.
+
+Upon termination, you must cease all use of the Software.
+
+## 8. Disclaimer of Warranties
+
+THE SOFTWARE IS PROVIDED "AS IS" AND "AS AVAILABLE", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT.
+
+## 9. Limitation of Liability
+
+TO THE MAXIMUM EXTENT PERMITTED BY LAW, IN NO EVENT SHALL NUTRIENT BE LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, OR EXEMPLARY DAMAGES, INCLUDING BUT NOT LIMITED TO LOSS OF PROFITS, DATA, OR USE, ARISING OUT OF OR RELATED TO THE SOFTWARE.
+
+## 10. Governing Law
+
+This License shall be governed by and construed in accordance with the laws of the State of North Carolina, without regard to conflict of law principles.
+
+## 11. Usage Data
+
+The Software may collect and transmit usage data related to performance, feature usage, and document processing activity. This data does not include the contents of documents or personally identifiable information.
+
+By using the Software, you agree to the collection and use of such data in accordance with Nutrient's Privacy Policy.
+
+## 12. Third-Party Software
+
+This Software incorporates third-party open source components. A full list of acknowledgements is available at: https://www.nutrient.io/legal/acknowledgements/nutrient-cli-acknowledgements/

README.md

@@ -0,0 +1,172 @@
+# Nutrient PDF to Markdown
+
+[![License: Proprietary](https://img.shields.io/badge/license-Nutrient_Free_Use-blue)](LICENSE.md)
+[![macOS](https://img.shields.io/badge/macOS-arm64-brightgreen)](https://github.com/PSPDFKit/pdf-to-markdown)
+[![Linux](https://img.shields.io/badge/Linux-x64_|_arm64-brightgreen)](https://github.com/PSPDFKit/pdf-to-markdown)
+[![Windows](https://img.shields.io/badge/Windows-x64_(coming_soon)-yellow)](https://github.com/PSPDFKit/pdf-to-markdown)
+
+<p align="center">
+  <img src="docs/assets/demo.gif" alt="pdf-to-markdown demo" width="720">
+</p>
+
+**Stop wasting your context window on PDF extraction.**
+
+Fast, accurate Markdown from PDFs — locally, with no cleanup required. Built for Claude, Codex, RAG pipelines, and document-heavy automation where noisy extraction burns tokens and makes downstream results less reliable.
+
+- **How fast is it?** — 0.007s per page. 90x faster than docling, 37x faster than pymupdf4llm. ([benchmarks](#benchmarks))
+- **How accurate is it?** — 0.92 reading order (best in class), 0.88 overall extraction accuracy, 0.81 heading detection. ([benchmarks](#benchmarks))
+- **Where do my PDFs go?** — Nowhere. The CLI runs locally. Your documents are not uploaded to Nutrient. ([trust & licensing](#trust-and-licensing))
+- **What does it cost?** — Free for up to 1,000 documents per calendar month. No license key, no signup, no API token. ([license](LICENSE.md))
+
+## Install
+
+### Agent skill (recommended)
+
+If you use Claude Code, Codex, Pi, Cursor, or Gemini CLI, install the [Nutrient Skills](https://github.com/pspdfkit-labs/nutrient-skills) plugin — the extraction runs automatically when your agent needs to read a PDF:
+
+```bash
+npx skills add pspdfkit-labs/nutrient-skills --skill pdf-to-markdown
+```
+
+Or with marketplace/plugin flows (Claude Code, Codex):
+
+```text
+/plugin marketplace add pspdfkit-labs/nutrient-skills
+/plugin install pdf-to-markdown@nutrient-skills
+```
+
+With Pi:
+
+```bash
+pi install git:github.com/PSPDFKit-labs/nutrient-skills
+```
+
+Once installed, just reference a PDF in your prompt — no extra commands needed:
+
+> "Extract the pricing table from proposal.pdf"
+
+The skill invokes the CLI transparently and passes the resulting Markdown into your agent context.
+
+### Standalone CLI
+
+For use outside an agent, install the CLI directly:
+
+```bash
+curl -fsSL https://raw.githubusercontent.com/PSPDFKit/pdf-to-markdown/main/install.sh | sh
+```
+
+This installs `pdf-to-markdown` into `~/.local/bin` by default.
+
+You can also install from a clone:
+
+```bash
+git clone https://github.com/PSPDFKit/pdf-to-markdown.git
+cd pdf-to-markdown
+./install.sh            # or: npm install -g .
+```
+
+## Usage
+
+### Single PDF
+
+```bash
+pdf-to-markdown input.pdf output.md
+```
+
+If `output.md` is omitted, Markdown is written to stdout.
+
+### Batch directory
+
+```bash
+pdf-to-markdown ./input-pdfs ./output-markdown
+```
+
+When both arguments are directories, the CLI converts every PDF in the input directory and writes matching Markdown files into the output directory.
+
+## Platform Support
+
+- macOS Apple Silicon (`Darwin/arm64`)
+- Linux x86_64
+- Linux arm64
+- Windows x64 (coming soon)
+
+## Benchmarks
+
+Benchmark results from 200 PDF documents with hand-annotated Markdown ground truth, evaluated using NID (reading order), TEDS (table structure), and MHS (heading hierarchy) metrics. Benchmarked on `2026-04-02`.
+
+### Visual Snapshot
+
+![Extraction accuracy](docs/assets/extraction-accuracy.png)
+
+![Reading order](docs/assets/reading-order.png)
+
+![Table structure](docs/assets/table-structure.png)
+
+![Heading level](docs/assets/heading-level.png)
+
+![Extraction speed](docs/assets/extraction-speed.png)
+
+![Faster with Nutrient](docs/assets/faster-with-nutrient.png)
+
+### Accuracy
+
+| Solution | Overall | Reading Order (NID) | Table Structure (TEDS) | Heading Level (MHS) |
+| --- | ---: | ---: | ---: | ---: |
+| docling | **0.88** | 0.90 | **0.89** | **0.82** |
+| **Nutrient** | **0.88** | **0.92** | 0.66 | 0.81 |
+| opendataloader | 0.83 | 0.90 | 0.49 | 0.74 |
+| pymupdf4llm | 0.83 | 0.88 | 0.48 | 0.78 |
+| markitdown | 0.59 | 0.84 | 0.27 | 0.00 |
+| pypdf | 0.58 | 0.87 | 0.00 | 0.00 |
+| liteparse | 0.57 | 0.86 | 0.00 | 0.00 |
+
+### Speed
+
+| Solution | Seconds per page |
+| --- | ---: |
+| **Nutrient** | **0.007** |
+| opendataloader | 0.014 |
+| pypdf | 0.019 |
+| markitdown | 0.106 |
+| liteparse | 0.233 |
+| pymupdf4llm | 0.252 |
+| docling | 0.618 |
+
+### Faster with Nutrient
+
+- `90x` faster than `docling`
+- `37x` faster than `pymupdf4llm`
+- `34x` faster than `liteparse`
+- `15x` faster than `markitdown`
+- `3x` faster than `pypdf`
+- `2x` faster than `opendataloader`
+
+For the full comparison table, see [docs/benchmarks.md](docs/benchmarks.md).
+
+## Trust and Licensing
+
+- Free for up to `1,000` documents per calendar month
+- PDFs stay local — your documents are not uploaded to Nutrient by this extractor
+- A commercial license is required for processing more than `1,000` documents per month
+- The extraction engine is delivered as a signed platform binary; the repo contains only the wrapper and documentation
+- The license is non-transferable — you may not redistribute the binary standalone or sublicense it to third parties; embedding it in your own application is permitted under the free tier terms
+
+See [LICENSE.md](LICENSE.md) for the full terms and [docs/distribution-model.md](docs/distribution-model.md) for details on what ships in this repo vs. the binary.
+
+## FAQ
+
+### What makes this different from other PDF extractors?
+
+Speed and accuracy should not be a tradeoff. Most extractors are either fast but lose structure (markitdown, pymupdf4llm) or accurate but slow (docling). Nutrient extracts at 0.007s per page with strong reading order, heading, and table preservation — less cleanup, fewer wasted tokens, and more reliable downstream results.
+
+### Do my documents leave my machine?
+
+No. The CLI processes PDFs locally. Nothing is uploaded to Nutrient. Note that if you feed the extracted Markdown into Claude, Codex, or another model provider, their own data policies apply.
+
+### Do I need a license key or API token?
+
+No. There is no signup, no license key, and no API token. Install the CLI and start converting. The free tier (up to 1,000 documents per calendar month) is enforced via the [license terms](LICENSE.md), not a technical gate. If you need to process more than 1,000 documents per month, contact `sales@nutrient.io` for a commercial license.
+
+### Why is the extraction engine closed-source?
+
+The repo is designed to be reviewable — you can read the wrapper, the installer, and the documentation. The extraction engine is distributed as a signed binary to protect the implementation while keeping the CLI surface fully transparent.