ltspice-mcp
MCP server for automating LTspice on macOS, enabling simulation, schematic generation, data extraction, verification, and rendering via natural language or agents.
README
LTspice MCP for macOS
Model Context Protocol (MCP) server that lets agents and MCP clients control LTspice on macOS for simulation, schematic generation, data extraction, verification, and rendering.
This project is designed for practical automation: reliable runs, reproducible artifacts, and outputs that match LTspice behavior closely.
Inspired by:
What You Can Do
- Run LTspice simulations from MCP (
simulateNetlistFile,runSimulation, queue tools). - Generate and refine schematics (
createSchematic*, lint/clean/debug tools). - Render real LTspice images for schematics, plots, and symbols.
- Query RAW vectors and analysis metrics (bandwidth, margins, rise/fall, settling).
- Automate
.measand assertion-driven verification workflows. - Run stepped and Monte Carlo studies with structured results.
Quick Start (5 Minutes)
1) Prerequisites
- macOS with LTspice installed (
/Applications/LTspice.appexpected by default). - Python 3.11+.
uv(recommended) orpip.
2) Install
uv sync
If you prefer pip:
python3 -m venv .venv
source .venv/bin/activate
pip install -e .
3) Start the MCP daemon
./scripts/ltspice_mcp_daemon.sh start
./scripts/ltspice_mcp_daemon.sh status
Default endpoint:
http://127.0.0.1:8765/mcp
4) Grant macOS permissions once
Required for screenshot/render and some UI automation features.
./scripts/ltspice_mcp_daemon.sh trigger-initial-permissions
./scripts/ltspice_mcp_daemon.sh check-accessibility
5) Run a smoke test
uv run python smoke_test_mcp.py \
--transport streamable-http \
--server-url http://127.0.0.1:8765/mcp
Client Configuration
URL-capable MCP clients
[mcp_servers.ltspice]
url = "http://127.0.0.1:8765/mcp"
enabled = true
Claude Desktop via mcp-remote
{
"mcpServers": {
"ltspice-mcp": {
"command": "/opt/homebrew/bin/npx",
"args": ["-y", "mcp-remote", "http://127.0.0.1:8765/mcp"]
}
}
}
stdio (subprocess) mode
{
"mcpServers": {
"ltspice-mcp": {
"command": "ltspice-mcp",
"args": ["--transport", "stdio"],
"cwd": "/absolute/path/to/ltspice-mcp"
}
}
}
Core Capabilities by Category
Setup and diagnostics
getLtspiceStatus,getLtspiceUiStatusdaemonDoctor,tailDaemonLog,getRecentErrors,getCaptureHealth
Simulation and queueing
simulateNetlist,simulateNetlistFile,runSimulationqueueSimulationJob,listJobs,jobStatus,cancelJob,listJobHistory
Schematic workflows
createSchematic,createSchematicFromNetlist,createSchematicFromTemplatevalidateSchematic,lintSchematic,autoDebugSchematicinspectSchematicVisualQuality,autoCleanSchematicLayout
Data, measurements, and verification
getPlotNames,getVectorsInfo,getVectorData,getLocalExtremagetBandwidth,getGainPhaseMargin,getRiseFallTime,getSettlingTimeparseMeasResults,runMeasAutomation,runVerificationPlan,runSweepStudy
Native LTspice rendering
renderLtspiceSchematicImagerenderLtspicePlotImage,renderLtspicePlotPresetImagerenderLtspiceSymbolImagestartLtspiceRenderSession,endLtspiceRenderSession
Reliability Notes
- Streamable HTTP defaults are tuned for compatibility:
json_response = truestateless_http = true
- UI integration is disabled by default (
LTSPICE_MCP_UI_ENABLED=0). - Schematic single-window updates are enabled by default.
- Rendering uses LTspice + ScreenCaptureKit direct-window capture.
- Run artifacts are stored per run to keep historical results stable.
Daemon Operations
./scripts/ltspice_mcp_daemon.sh start
./scripts/ltspice_mcp_daemon.sh restart
./scripts/ltspice_mcp_daemon.sh stop
./scripts/ltspice_mcp_daemon.sh status
./scripts/ltspice_mcp_daemon.sh logs --lines 200
./scripts/ltspice_mcp_daemon.sh logs --follow
Permission helpers:
./scripts/ltspice_mcp_daemon.sh trigger-screen-recording-permission
./scripts/ltspice_mcp_daemon.sh trigger-accessibility-permission
Testing
Run core tests:
PYTHONPATH=src .venv/bin/python -m unittest discover -s tests -v
Run real ScreenCaptureKit integration tests (opt-in):
LTSPICE_MCP_RUN_REAL_SCK=1 PYTHONPATH=src .venv/bin/python -m unittest tests.test_screencapturekit_integration -v
LTSPICE_MCP_RUN_REAL_SCK=1 PYTHONPATH=src .venv/bin/python -m unittest tests.test_plot_render_mcp_real -v
Contributing
Contributions are welcome. Start with:
- CONTRIBUTING.md
- AGENT_README.md for agent-specific workflows
When filing bugs, include:
- MCP server version,
- LTspice version,
- transport mode,
- exact tool call + parameters,
- daemon log excerpts.
Documentation Map
- docs/README.md
- AGENT_README.md
- CHANGELOG.md
- COMPATIBILITY.md
- CONTRIBUTING.md
- SUPPORT.md
- CODE_OF_CONDUCT.md
- SECURITY.md
- MCP resource:
docs://agent-readme - MCP tool:
readAgentGuide
Important Environment Variables
Core:
LTSPICE_BINARYLTSPICE_MCP_WORKDIRLTSPICE_MCP_TIMEOUTLTSPICE_MCP_TRANSPORTLTSPICE_MCP_HOSTLTSPICE_MCP_PORTLTSPICE_MCP_STREAMABLE_HTTP_PATHLTSPICE_MCP_JSON_RESPONSELTSPICE_MCP_STATELESS_HTTP
UI/render:
LTSPICE_MCP_UI_ENABLEDLTSPICE_MCP_SCHEMATIC_SINGLE_WINDOWLTSPICE_MCP_SCHEMATIC_LIVE_PATHLTSPICE_MCP_SCK_HELPER_DIRLTSPICE_MCP_SCK_HELPER_PATHLTSPICE_MCP_VERIFY_WINDOW_CLOSE
Logging:
LTSPICE_MCP_LOG_LEVELLTSPICE_MCP_TOOL_LOGGINGLTSPICE_MCP_TOOL_LOG_MAX_ITEMSLTSPICE_MCP_TOOL_LOG_MAX_CHARSLTSPICE_MCP_DISABLE_UVICORN_NOISE_FILTERSLTSPICE_MCP_DAEMON_LOG_DIR
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。