KiCad Netlist Tool

Tool for reducing LLM token count from Kicad SCH files, to allow for automated documentation

PythonKiCadAI

Compact netlist extraction for LLM workflows

The Problem

KiCad schematic files are mostly noise for LLM purposes. A modest circuit is 55KB and 25,000+ tokens. Most of that is UUIDs, coordinates, symbol graphics, and metadata. The actual electrical information is maybe 2% of the file.

The Tool

Extracts just the electrical information and outputs it in TOKN format:

Original: 1,509,037 tokens → TOKN: 51,443 tokens
Reduction: 96.6%

KiCad Netlist Tool

Features:

  • Hierarchical sheet selection — pick which sheets to include
  • Copy to clipboard — paste directly into your LLM
  • File monitoring — auto-regenerate when you save in KiCad
  • Real-time stats — token count before and after

TOKN Format

The output is TOKN v1.2 — a compact format that preserves the electrical information:

# TOKN v1
title: Audio Preamp

components[3]{ref,type,value,fp,x,y,w,h,a}:
  U1,ECC83,ECC83-1,Valve,127.00,85.09,25.40,20.32,0
  R1,R,1.5k,0805,149.86,85.09,7.62,0.00,90
  C1,C,10uF,RadialD10,123.19,64.77,0.00,7.62,0

nets[2]{name,pins}:
  VIN,U1.2,C1.1
  VOUT,U1.7,R1.2

Components, values, footprints, positions, and net connectivity.

Use Cases

Documentation: Generate functional descriptions of circuit blocks.

Design review: Catch missing pull-ups, floating pins, etc.

BOM generation: Component list is already extracted.

Selective extraction: Include only the sheets you need.

Tech Stack

  • Language: Python
  • GUI: CustomTkinter
  • Format: TOKN v1.2
  • Platforms: Windows, macOS, Linux

Source Code: GitHub