DocDot · Unified Local Parser

The Simplest Way to RunAny PDF Parser, Locally

One-line install for local PDF parsing on macOS.

Select different parsers as you need, and get upstream-quality output with zero cloud upload.

*Currently only runs on M-Chip Macs. Linux and Windows coming soon.

A Complete Parser Stack for Agent Workflows

Access all local PDF parsers for macOS in one place, keep up with new releases, and use the output directly with your agents.

1
Install DocDot
As prompted in the terminal, you may need to run source ~/.docdot/env manually to apply the PATH configuration.
curl -fsSL https://docdot.ai/install.sh | bash
2
Install the Parser
Currently supportsMinerUPaddleOCRGLM-OCRLiteParseDocling, and stays updated with the latest PDF parsers.
docdot install mineru
3
Parse the PDF
Parse the PDF using this command
docdot -i ./example.pdf -o ./output
Or parse the PDF using web UI
docdot web #Then open http://127.0.0.1:8173 in your browser.
4
Install DocDot Agent Skill
Currently supportsCodexClaude CodeOpenClawHermesWorkBuddy
docdot skills install
Restart the Agent tool upon Skills exceptions.(Codex: The DocDot Agent Skill is only supported inside Codex Projects.)

Find the Right PDF Parser

Note: Data are as of Jul 29, 2026.

Step 1. Filter Before You Shortlist

Shortlist

5 tools found3 selected

MinerU

3.4.4

Broad language support with balanced benchmark coverage.

Overall73.35↑

Overall Edit0.34↓

FPS0.33↑

Avg Power8.90 W↓

PaddleOCR

3.7.0

Wide language coverage and solid general-purpose OCR performance.

Overall71.63↑

Overall Edit0.27↓

FPS0.06↑

Avg Power7.16 W↓

Docling

2.122.0

High-fidelity, multi-format parser tailored for AI pipelines.

Overall67.75↑

Overall Edit0.33↓

FPS0.94↑

Avg Power15.29 W↓

GLM-OCR

0.1.5

Strong on mixed-layout extraction with chart support.

Overall65.74↑

Overall Edit0.35↓

FPS0.12↑

Avg Power6.75 W↓

LiteParse

2.5.1

Efficiency-first parser with lower structural fidelity.

Overall26.98↑

Overall Edit0.72↓

FPS5.14↑

Avg Power8.13 W↓

More providers will be available

Step 2. Compare Features, Accuracy, and Runtime.

MinerU
PaddleOCR
Docling

Arrows indicate whether higher or lower values are preferable for each metric.

Features	MinerU	PaddleOCR	Docling
Output Format	markdown, json	markdown, json	markdown, json
Scanned Docs
Multilingual	50+ Langs	50+ Langs	50+Langs
Layout Coordinates
Tables
Cross-Page Table
Formulas
Image Extraction
Charts
Size on Disk	3.21 GiB	2.57 GiB	3.5 GiB

The Simplest Way to RunAny PDF Parser, Locally

A Complete Parser Stack for Agent Workflows

Install DocDot

Install the Parser

Parse the PDF

Install DocDot Agent Skill

Find the Right PDF Parser

Step 1. Filter Before You Shortlist

MinerU

PaddleOCR

Docling

GLM-OCR

LiteParse

Step 2. Compare Features, Accuracy, and Runtime.