By using this site, you agree to the Privacy Policy and Terms of Use.
Accept
Logic & LayersLogic & Layers
  • Tools
  • Earn with AI
  • Productivity
  • Automation
  • Guides
Logic & LayersLogic & Layers
  • Privacy Policy
  • About
Search
  • Tools
  • Earn with AI
  • Productivity
  • Automation
  • Guides
  • About
  • Contact
  • Blog
  • Privacy Policy
  • Complaint
  • Advertise
© 2026 Logic and Layers. Ruby Design Company. All Rights Reserved.
PCB design with Claude AI using KiCad-MCP
Automation

PCB Design with Claude: AI-Assisted Hardware Design for Beginners

Editorial Team
Last updated: August 23, 2026 9:30 am
Editorial Team
Share
Claude designing PCB in KiCad via MCP

PCB design with Claude is now practical, not theoretical. An open-source bridge called KiCad-MCP connects Claude directly to KiCad’s PCBnew API. Describe the circuit in plain English. Claude writes the Python code that drives KiCad. The board appears on screen. Iterate by talking. That’s it.

Contents
Why PCB Design Was Always HardEnter KiCad-MCP: The Bridge That Changes EverythingWhat You Need Before StartingStep 1: Install KiCad-MCP ServerStep 2: Configure Claude DesktopStep 3: Create Your First Schematic with AIStep 4: Lay Out the Board Without HeadachesStep 5: Export for ManufacturingPro Tips for Better ResultsTakeaway: Your First PCB Is One Conversation Away

Testing PCB design with Claude using zero recent KiCad experience — my last board was a college project fifteen years ago — produced a working ESP32 sensor board in forty-five minutes. The Gerbers checked out on the first try.

Let me show the exact setup for your first board.

Why PCB Design Was Always Hard

Traditional PCB design forces CAD-tool thinking. Components placed by coordinates. Traces routed by hand, nudged millimeter by millimeter. Symbol libraries, footprint libraries, 3D model libraries managed separately — each with unique quirks and naming conventions.

The schematic editor demands knowing the difference between a net label and a global label. The layout editor punishes missing design rules before routing starts. The footprint wizard is its own nightmare.

Most beginners quit at “associate footprints.” Happened to me twice.

Complexity isn’t the problem. The tools expose every knob and lever simultaneously, with zero guardrails. You’re not designing a circuit. You’re managing a database of geometric constraints.

Enter KiCad-MCP: The Bridge That Changes Everything

KiCad-MCP is an open-source Model Context Protocol server sitting between Claude and KiCad. It exposes KiCad’s Python API as tools Claude can call. Say “add a 10k resistor to the schematic.” Claude calls the MCP tool. The resistor appears.

The project lives on GitHub (kicad-mcp/kicad-mcp). Free, open source, actively maintained. Clean architecture: TypeScript MCP server handles protocol, modular Python backend does board manipulation.

Understanding MCP isn’t required. Installation is.

What You Need Before Starting

Hardware and software requirements:
– KiCad 8.0 or newer (free from kicad.org)
– Claude Desktop app with Pro or Max subscription (free web version can’t run MCP servers)
– Git and Node.js installed (for building the MCP server)
– A GitHub account (to clone the repo)

Mental prerequisites:
– Basic electronics knowledge (resistor, capacitor, microcontroller functions)
– Willingness to read error messages and retry
– Patience — bleeding-edge stuff breaks

That’s the list. KiCad’s interface stays unknown. Python skills stay unnecessary. PCB design with Claude handles the code for you.

Step 1: Install KiCad-MCP Server

Open your terminal. Clone the repository:

git clone https://github.com/kicad-mcp/kicad-mcp.git
cd kicad-mcp

Install dependencies and build:

npm install
npm run build

Compilation takes thirty seconds. “Build complete” without red errors means you’re ready.

Step 2: Configure Claude Desktop

This step traps most people. Claude Desktop reads MCP configuration from a JSON file. Location depends on OS:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json

Create the file if missing. Add this configuration:

{
  "mcpServers": {
    "kicad": {
      "command": "node",
      "args": ["/full/path/to/kicad-mcp/dist/index.js"],
      "env": {
        "KICAD_PYTHON_PATH": "/Applications/KiCad/KiCad.app/Contents/Frameworks/Python.framework/Versions/Current/bin/python3"
      }
    }
  }
}

Replace /full/path/to/kicad-mcp with your actual clone path. The KICAD_PYTHON_PATH must point to KiCad’s embedded Python — critical detail. Windows typically uses C:\Program Files\KiCad\8.0\bin\python.exe. Linux typically uses /usr/bin/python3.

Restart Claude Desktop completely. Quit and reopen. A small hammer icon in the chat input’s bottom-left corner confirms MCP connection.

Step 3: Create Your First Schematic with AI

Start a new chat in Claude. Say this exactly:

“I want to design a simple PCB. Create a new KiCad project called ‘esp32-temp-sensor’ in my Documents folder. Then add an ESP32-DevKitC-V4, a DS18B20 temperature sensor, a 4.7k pull-up resistor, and a 0.1µF decoupling capacitor. Connect the sensor data pin to GPIO 4, the resistor between data and 3.3V, and the capacitor across the sensor’s power pins.”

Watch what happens. Claude will:
1. Create the project and schematic file
2. Search symbol libraries for each component
3. Place them on the schematic canvas
4. Wire connections using net labels
5. Annotate reference designators (U1, R1, C1, etc.)

The schematic appears in KiCad’s window in real time. Weirdly satisfying.

Pro tip: Missing symbol? Claude tells you. Say “use a generic 8-pin MCU symbol for the ESP32” or “create a custom symbol for the DS18B20.” Both work.

Step 4: Lay Out the Board Without Headaches

Now the fun part. Say:

“Switch to the PCB editor. Set the board outline to a rounded rectangle, 60mm by 40mm, with 2mm corner radius. Place the ESP32 module near the left edge, USB connector facing out. Put the sensor near the right edge. Route traces with 0.2mm width, 0.2mm clearance. Add a ground pour on both layers.”

Claude will:
1. Create the board outline
2. Place components with reasonable orientation
3. Run the auto-router (or manual routing via API calls)
4. Add copper pours connected to GND net
5. Generate 3D view

Iterate freely: “Move the sensor closer to the edge.” “Make the USB connector overhang by 2mm.” “Add mounting holes at four corners, 3mm diameter.”

Each instruction triggers a new API call. The board updates live.

Step 5: Export for Manufacturing

Once satisfied, say:

“Generate Gerber files and drill files for JLCPCB. Use the standard 2-layer stackup. Include the edge cuts layer. Zip everything up.”

Claude runs plot and drill commands, packages files, tells you the location. Upload to JLCPCB, PCBWay, or preferred fab. Five boards cost $2.

Two weeks later, solder components. It works.

Pro Tips for Better Results

Use version control. Initialize a git repo in your project folder. Commit after each major step. Claude makes mistakes — sometimes deletes traces it just routed. Git rolls back instantly.

Be specific about constraints. “Route with 0.2mm traces” beats “route the board.” “Keep the antenna area clear” prevents non-functional WiFi boards.

Use the 3D view. “Show me the 3D render” after placement catches mechanical conflicts (USB connector hitting enclosure, tall components blocking each other) before fabrication.

Save your prompts. Working prompts become a personal PCB design library. Next board takes ten minutes.

Don’t trust everything. Always run KiCad’s DRC (Design Rule Check) before exporting. Claude misses things — unconnected nets, overlapping courtyards, missing keepout zones. DRC catches them.

Takeaway: Your First PCB Is One Conversation Away

PCB design used to demand months of learning. Now it’s an afternoon conversation. The tools haven’t changed — KiCad remains KiCad. The interface has. Talk. Code executes. Boards appear.

Start simple. Blinky LED board. Sensor breakout. Power supply module. Each teaches vocabulary for the next.

When weird errors hit — “footprint not found,” “net class not defined,” “zone fill failed” — paste the error into Claude. It fixes them. That’s the real superpower: not designing boards, but debugging them together.

Go make something. The fab house is waiting.

You Might Also Like

Relay.app is shutting down: best alternatives to move to
Zapier Tables explained: Automate your data without spreadsheets
ChatGPT wallet: OpenAI is building a way for AI to spend your money
How much does AI customer support actually save?
Firecrawl Monitor: Let AI watch the web for you
TAGGED:AI tutorialClaude AIhardware designKiCadPCB design
Share
Previous Article How to Protect Your AI Content: Bypassing Claude’s Watermarks and Staying Safe
Next Article Claude Code terminal showing KiCad MCP server generating PCB layout Claude Code for KiCad: Automate PCB Design with AI Scripts
Leave a Comment

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

banner banner
Create an Amazing Newspaper
Discover thousands of options, easy to customize layouts, one-click to import demo and much more.
Learn More

Latest News

The Most Customizable LLM Chat App Is Free and Open Source (Setup Guide)
Tools
Hooded hacker figure with the OpenAI logo as a face, surrounded by panicked emoji faces on a blue and orange background
OpenAI’s Hugging Face Hack: What It Means for Your AI Safety
Tools
Smartphone displaying the Claude app logo with the orange Anthropic starburst icon on a black background
Claude Code Session Messaging + Auto Mode: Complete Beginner Guide
Productivity
How Headway Built Custom AI Tool with Claude Code SDK
Guides

Recent Posts

  • The Most Customizable LLM Chat App Is Free and Open Source (Setup Guide)
  • OpenAI’s Hugging Face Hack: What It Means for Your AI Safety
  • Claude Code Session Messaging + Auto Mode: Complete Beginner Guide
  • How Headway Built Custom AI Tool with Claude Code SDK
  • Gemini Chrome Select from Screen: Beginner Guide

Recent Comments

  1. I tested 6 AI task managers for 30 days (Only 3 are worth it) on Best AI time blocking apps in 2026 (I tested 5 that survive when your schedule falls apart)
  2. Gemini CLI: How to Start Coding with AI for Free on How to use Google Gemini 3.5 Flash Search: A complete beginner guide
  3. GitHub Copilot's New Pricing: 10x More Expensive? | Logic & Layers on Cancel ChatGPT, Perplexity & Gemini — use Claude instead
  4. Google Gemini Spark Review: Is It Worth Using? | Logic & Layers on Gemini in Android Auto: Complete beginner’s guide (2026)
  5. Google Gemini Spark Review: Is It Worth Using? | Logic & Layers on Cancel ChatGPT, Perplexity & Gemini — use Claude instead

You Might also Like

AI automation framework - 4 roles of AI in workflows
Automation

The 4 Roles AI Plays in Automation (And When to Skip AI Entirely)

Editorial Team
Editorial Team
9 Min Read
AI agent safety concept showing sandboxed environment
Automation

AI Agent Safety for Beginners: Run Agents Without Risk

Editorial Team
Editorial Team
9 Min Read
Abstract representation of AI model routing between premium and free tiers
Tools

ChatGPT has been secretly routing Pro to mini

Editorial Team
Editorial Team
9 Min Read
//

We influence 20 million users and is the number one business and technology news network on the planet

Quick Link

  • PRIVACY NOTICE
  • YOUR PRIVACY RIGHTS
  • INTEREST-BASE ADSNew
  • TERMS OF USE
  • OUR SITE MAP

Support

  • ADVERTISE
  • ONLINE BESTHot
  • CUSTOMER
  • SERVICES
  • SUBSCRIBE

Categories

  • Tools
© 2026 Logic and Layers. All Rights Reserved.