Getting Started with GerberGPT

Quick Overview

What is GerberGPT?

GerberGPT is an AI platform that turns natural-language specs into schematics and PCB layouts. It’s tuned on electronics engineering data and has access to a large component database, so it can generate full designs from your text requirements.

What is MCP?

The Model Context Protocol (MCP) is an open protocol (from Anthropic) that lets AI agents talk to external tools and services in a standardized way, usually over JSON-RPC via stdio or SSE. In Trae IDE, MCP is used to connect the AI agent to things like databases, APIs, or custom services.

What is Trae IDE?

Trae is an AI-powered IDE that includes an integrated agent, multi-model support, and first-class MCP integration starting from Trae IDE v1.3.0. It reads MCP configuration from JSON files and .rules files to control agent behavior.

2. Why use MCP with GerberGPT?

Using MCP as the bridge between Trae and GerberGPT has several benefits:

Standard, reusable integration

You define GerberGPT once as an MCP server and any MCP-compatible client (Trae, other IDEs, desktop clients) can use it with the same config.

Tighter PCB workflows inside the IDE

The Trae agent can read your project files, use MCP to call GerberGPT to propose PCB designs, then write code or documentation back into your repo — all from one chat.

Separation of concerns & security

API keys and endpoints for GerberGPT live in MCP configuration, not in prompts. That’s cleaner and safer for team projects.

Better long-term behavior with .rules

Trae’s .rules files let you describe how the agent should collaborate with GerberGPT (e.g. always validate footprints, follow your house design rules), and those rules get loaded automatically by the agent.

Future-proofing

If GerberGPT improves or adds new tools (e.g. BOM optimization, DFM checks), they can be exposed via MCP without changing your core Trae setup.

Prerequisites

Before you start, make sure you have:

Trae IDE v1.3.0 or later

– This is the version that ships MCP support and .rules.

A GerberGPT account and access to the MCP client/server info

– Follow the official “How to configure GerberGPT MCP client” tutorial on GerberGPT’s site to get:

Your API key or auth token

The MCP server command (for stdio) or URL (for SSE)

Basic familiarity with:

Editing JSON files

Your OS terminal (if using a local stdio MCP server)

Step 2 – Get GerberGPT MCP connection details

From the GerberGPT side:

Go to the “How to configure GerberGPT MCP client” tutorial or developer/settings page.

Note down:

The recommended MCP connection type:

stdio command (e.g. a CLI you run locally), or

sse URL (e.g. https://.../mcp)

Any required environment variables (e.g. GERBERGPT_API_KEY

Step 3 – Configure

mcp.json

in Trae (MCP client setup)

Trae supports both global and project MCP config; we’ll focus on project-level:

Project-level MCP config file:

.trae/mcp.json (recommended, keeps PCB config local to your repo)

Create or open .trae/mcp.json and add one of the following, depending on how GerberGPT exposes MCP:

Option A – GerberGPT via

stdio

(local CLI)

Use this if GerberGPT provides a CLI you run on your machine:

1{
2  "mcpServers": {
3    "gerbergpt": {
4      "url": "https://mcp.gerbergpt.com/mcp",
5      "headers": {
6        "token": "sk-**********"
7      }
8    }
9  }
10}

name: A short identifier; this is what Trae will show for the MCP server.

command: The CLI command (and args, if any) that starts the MCP server.

env: Environment variables required by GerberGPT (replace with real names).

Option B – GerberGPT via SSE (remote HTTPS endpoint)

Use this if GerberGPT hosts a remote MCP endpoint you call over SSE:

Step 4 – Add

.rules

so the agent knows how to use GerberGPT

Trae uses .rules markdown files to provide long-term context and behavior guidelines.

.trae/project_rules.md

10. Step 6 – Use GerberGPT from inside Trae

Now you can start using GerberGPT directly in your IDE with natural language:

Open the AI chat sidebar in Trae.

Select an agent that has MCP enabled (e.g. @Builder with MCP or your custom agent).

In your prompt, clearly describe the PCB task and mention you want it done via GerberGPT, for example:

“Using the gerbergpt MCP server, design a 2-layer PCB for an ESP32 dev board with USB-C power, onboard regulator, and broken-out GPIO headers. Target 5V input, 3.3V logic, max 1A load.”

Ask the agent to:

Generate or refine schematics / layout ideas

Propose component choices using GerberGPT’s component knowledge

Export or verify Gerber outputs for your fabrication house

Review the response:

Inspect any generated design files in your repo

Iterate in chat: “Tighten clearances for 4-layer stackup”, “swap to only JLCPCB basic parts”, etc.

11. Optional: Shareable tutorial structure (for your web page)

If you’re turning this into a full tutorial page, a clean outline could be:

Intro

Short description of GerberGPT, MCP, and Trae

Why MCP + GerberGPT?

Benefits table or bullet list

Architecture Overview

Small diagram showing Trae ⇄ MCP ⇄ GerberGPT

Prerequisites

Set Up MCP in Trae

Create .trae/mcp.json

stdio vs sse examples

Add Project Rules

Sample .trae/project_rules.md

Run and Test

Restart Trae, open AI panel, send your first PCB prompt

Next Steps

Hook in CI, share configs with teammates, experiment with advanced GerberGPT tools

You can copy-paste the sections above into your web tutorial, dropping in screenshots of Trae’s UI and your own GerberGPT dashboard where appropriate.