JLC EDA MCP Server
Enables MCP clients to control JLCPCB EDA (嘉立创 EDA) for PCB and schematic design operations through a bridge and plugin architecture.
README
JLC EDA MCP Server
这是从原 jlc_import 目录整理出来的独立仓库,用来让支持 Model Context Protocol 的客户端控制嘉立创 EDA。
Credits
当前仓库基于 hyl64/jlcmcp 整理和改造。
git clone <this-repo-url> jlc-mcp
cd jlc-mcp
仓库包含三个部分:
src/: MCP server,向 Codex、Claude Code、Cursor、Windsurf 等客户端暴露 PCB/原理图工具。local_jlc_gateway.cjs: 本机 gateway,接收 MCP server 的 HTTP 命令,并通过 WebSocket 转发给 EDA 插件。jlc-bridge/: 嘉立创 EDA 扩展插件,运行在 EDA 里执行实际操作。
架构
MCP client
-> stdio
MCP server: <absolute-path-to-jlc-mcp>/dist/index.js
-> HTTP POST http://127.0.0.1:18800/command
local gateway: local_jlc_gateway.cjs
-> WebSocket ws://127.0.0.1:18800/ws/bridge
jlc-bridge extension
-> 嘉立创 EDA
MCP server 本身不直接控制 EDA,它只把工具调用发送给本机 gateway。jlc-bridge 插件启动后会连接 gateway 的 /ws/bridge,再由插件调用嘉立创 EDA 的扩展 API。
前置条件
- Node.js 18 或更新版本
- 嘉立创 EDA 专业版,能安装本地扩展
- 一个支持 MCP 的客户端
- 如需使用
pcb_agent,需要设置ANTHROPIC_API_KEY
安装
cd /path/to/jlc-mcp
npm install
npm run build
构建成功后,MCP 入口文件是:
/path/to/jlc-mcp/dist/index.js
构建并安装 EDA 插件
cd /path/to/jlc-mcp/jlc-bridge
npm install
npm run build
构建后会生成:
/path/to/jlc-mcp/jlc-bridge/build/jlc-bridge.eext
/path/to/jlc-mcp/jlc-bridge/build/jlc-bridge.lcex
在嘉立创 EDA 里安装其中一个扩展包。插件会在 EDA 启动后自动尝试连接:
ws://127.0.0.1:18800/ws/bridge
也可以在 EDA 的 JLC Bridge 菜单里查看状态或手动切换。
WebSocket 是默认通信方式。插件还保留文件轮询 fallback,默认目录名是 jlc-bridge;如果确实要使用文件轮询,可以在 EDA 开发者控制台里设置 localStorage.jlcBridgeDir 为本机可写目录后重启插件。
启动 gateway
在一个单独终端中运行:
cd /path/to/jlc-mcp
npm run gateway
默认监听:
http://127.0.0.1:18800
检查状态:
curl http://127.0.0.1:18800/state
如果 bridgeConnected 是 true,说明嘉立创 EDA 插件已经连上 gateway。
可选环境变量:
| 变量 | 默认值 | 说明 |
|---|---|---|
JLC_GATEWAY_HOST |
127.0.0.1 |
gateway 监听地址 |
JLC_GATEWAY_PORT |
18800 |
gateway 监听端口 |
配置 MCP 客户端
把下面配置加入你的 MCP 客户端配置文件。仓库里也有同样内容的 mcp_config.example.json。
{
"mcpServers": {
"jlceda": {
"command": "node",
"args": [
"/absolute/path/to/jlc-mcp/dist/index.js"
],
"env": {
"GATEWAY_HTTP_URL": "http://127.0.0.1:18800/command"
}
}
}
}
配置后重启 MCP 客户端。
如果要启用 pcb_agent,加上:
{
"env": {
"GATEWAY_HTTP_URL": "http://127.0.0.1:18800/command",
"ANTHROPIC_API_KEY": "sk-ant-...",
"AGENT_MODEL": "claude-sonnet-4-20250514"
}
}
运行顺序
- 启动 gateway。
- 打开嘉立创 EDA,并确认
jlc-bridge插件已连接。 - 启动或重启 MCP 客户端。
- 在客户端里调用
pcb_ping或pcb_get_state验证链路。
如果顺序反了也可以,插件和客户端可以稍后重连。但首次验证时按上面顺序更容易定位问题。
可用工具
当前 MCP server 注册 39 个工具,分组如下:
- 状态查询:
pcb_get_state、pcb_screenshot、pcb_run_drc、pcb_get_tracks、pcb_get_pads、pcb_get_net_primitives、pcb_get_board_info、pcb_get_feature_support、pcb_ping - 元件操作:
pcb_move_component、pcb_relocate_component、pcb_batch_move、pcb_select_component、pcb_delete_selected、pcb_create_component - 走线和过孔:
pcb_route_track、pcb_create_via、pcb_delete_tracks、pcb_delete_via - 铺铜和禁布区:
pcb_create_copper_pour、pcb_delete_pour、pcb_create_keepout、pcb_delete_keepout - 丝印:
pcb_get_silkscreens、pcb_move_silkscreen、pcb_auto_silkscreen - 高级规则:
pcb_create_diff_pair、pcb_list_diff_pairs、pcb_delete_diff_pair、pcb_create_equal_length、pcb_list_equal_lengths、pcb_delete_equal_length - 原理图:
sch_get_state、sch_get_netlist、sch_run_drc、pcb_open_document - 计算工具:
calc_impedance、calc_trace_width - Agent:
pcb_agent,仅在设置ANTHROPIC_API_KEY后注册
所有 PCB 坐标和尺寸参数默认使用 mil。
本地验证
只验证 MCP server 能启动并列出工具,不需要 gateway:
cd /path/to/jlc-mcp
npm run build
printf '%s\n' \
'{"jsonrpc":"2.0","id":0,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"manual","version":"1.0.0"}}}' \
'{"jsonrpc":"2.0","method":"notifications/initialized","params":{}}' \
'{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}' \
| node dist/index.js
端到端验证需要 gateway 和 EDA 插件都运行:
curl http://127.0.0.1:18800/state
然后在 MCP 客户端调用:
pcb_ping
pcb_get_state
项目结构
.
├── src/
│ ├── index.ts
│ ├── bridge-client.ts
│ ├── agent.ts
│ ├── calculators.ts
│ └── tools/
├── jlc-bridge/
│ ├── src/index.ts
│ ├── extension.json
│ ├── build/pack.js
│ ├── package.json
│ └── tsconfig.json
├── local_jlc_gateway.cjs
├── mcp_config.example.json
├── package.json
└── tsconfig.json
常见问题
jlc bridge is not connected
gateway 已启动,但 EDA 插件没有连上。检查嘉立创 EDA 是否已打开、插件是否安装并启用,然后访问:
curl http://127.0.0.1:18800/state
MCP 客户端找不到工具
先确认已经构建:
cd /path/to/jlc-mcp
npm run build
再确认 MCP 配置里的路径是当前仓库的 dist/index.js 绝对路径,不是旧的 jlc_import 或 /tmp/jlcmcp 路径。
端口被占用
改 gateway 端口:
JLC_GATEWAY_PORT=18801 npm run gateway
同时把 MCP 配置改成:
{
"GATEWAY_HTTP_URL": "http://127.0.0.1:18801/command"
}
注意:当前 jlc-bridge 插件源码里默认连接 ws://127.0.0.1:18800/ws/bridge。如果改 gateway 端口,也需要同步修改 jlc-bridge/src/index.ts 里的 WS_URL 后重新构建并安装插件。
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.