Show HN: SPICE simulation → oscilloscope → verification with Claude Code
TL;DR Highlight
This is an experimental case demonstrating that connecting a SPICE simulator and a real oscilloscope to Claude Code via an MCP server allows for creating a feedback loop where AI directly analyzes and verifies simulation results and actual waveform data.
Who Should Read
Firmware/hardware engineers who want to utilize LLMs in hardware design or embedded development, or developers curious about cases of connecting MCP servers to actual physical equipment.
Core Mechanics
- Simply 'describing circuit designs in natural language and having Claude create the circuit' works for simple circuits but quickly reaches its limits in complex designs. This is because the AI often fails to understand what is desired accurately.
- However, it was much more effective to adopt a method of connecting Claude Code to immediate feedback means (oscilloscope, SPICE simulator) as tools. Claude shines in an environment where feedback comes immediately.
- The biggest benefit in this workflow was data analysis. Claude automatically handles tedious tasks that were previously done visually, such as time axis normalization and data sorting.
- Two MCP servers were created: `lecroy-mcp` for controlling a LeCroy oscilloscope and `spicelib-mcp` wrapping a SPICE simulator library. This allows Claude to call the actual measurement equipment and simulator as tools.
- Important principle when connecting an oscilloscope: Since Claude cannot see the physical connection status, never let it guess what is connected where. You must manage to prevent old measurement data from remaining in the context.
- Do not paste raw measurement data directly into Claude's context, but save it to a file and have Claude indirectly read and process the file. This is a practical tip to prevent context contamination.
- When working with microcontrollers, explicitly provide the pinout/pinmux map and prepare a Makefile exposing functions like build/flash/ping/erase in advance to prevent Claude from improvising commands.
- This demo is configured with intentionally simple circuits like an RC filter, but the author explains that this approach scales well to much more complex circuits and real embedded projects.
Evidence
- "There were experiences trying to design boards with Claude Opus and SPICE simulations, but the board functionality was completely hallucinated, and it made absurd claims like 'discovering a secret $1 billion project'. Ultimately, none of the boards worked, and it was busy praising itself without realizing it had created gibberish.\n\nSimilarly, there was an opinion that attempting to automate circuit design/simulation/optimization was quite good at reading and explaining files, but could not properly modify designs beyond the simplest circuits. This is because it does not understand the physical concepts behind the circuit, and ultimately the agent was properly utilized only after defining the range of config files (step size, optimization algorithm, min/max, etc.) to be passed to the simulator in a script.\n\nA case of trying to automate KiCad schematic review was also shared, and the same wall was hit. Initially, having Claude directly grep/read the `.kicad_sch` file resulted in creating non-existent pin numbers, but the problem was mostly solved by generating JSON with a Python analyzer and having Claude read only the JSON. There were also technical questions about how `spicelib-mcp` handles models not in the basic library (whether it passes the `.lib` path as a tool arg or whether the server manages the registry directly). \n\nA practical tip was shared in the comments that using Mermaid diagrams makes it much less cumbersome to explain and modify circuits to LLMs.\n\nThere was also an opinion that local LLMs should be used instead of commercial services like Claude. The argument was that local LLMs are the future, even if they are slower, pointing out the frequent policy changes (rug-pull), price increases, and user-unfriendly direction of commercial services."
How to Apply
- If you have a lot of manual work comparing SPICE simulation results with actual oscilloscope measurements, you can connect `lecroy-mcp` and `spicelib-mcp` to Claude Code to automate waveform data analysis and alignment. This can significantly reduce the burden of repetitive preprocessing tasks like time axis normalization.
- To automate embedded firmware development with Claude Code, prepare a Makefile containing build/flash/ping/erase actions in advance and explicitly provide the pinout map to prevent Claude from creating commands on the fly and causing malfunctions.
- Pasting raw data from measurement equipment directly into Claude's context wastes tokens and contaminates the context. Instead, save the data to a file and configure Claude to access it indirectly via the file path to create a more stable analysis workflow.
- When connecting physical equipment to MCP, do not let Claude guess the physical connection status, but explicitly tell it what is connected to which channel at the start of the session. Also, periodically reset the session state to prevent previous measurement data from remaining in the context to obtain reliable results.
Terminology
Related Papers
Show HN: ctx – Search the coding agent history already on your machine
Claude Code, Cursor, Codex 등 코딩 에이전트가 이전 세션의 논의·결정·실패 시도를 잊지 않도록 SQLite로 인덱싱해 재사용할 수 있게 해주는 오픈소스 CLI 도구다.
Micro-Agent: Beat Frontier Models with Collaboration Inside Model API
vLLM 팀이 단일 모델 API 호출 뒤에서 여러 모델이 협업하는 'Micro-Agent' 개념을 공개했습니다. 별도의 에이전트 코드 없이 라우터 레이어에서 모델 조합을 실행해 GPT-4급 결과를 더 저렴하게 낼 수 있다는 아이디어입니다.
Ornith-1.0: self-improving open-source models for agentic coding
Gemma 4와 Qwen 3.5를 기반으로 파인튜닝한 코딩 특화 오픈소스 모델로, RL(강화학습)을 통해 스캐폴드(에이전트 실행 구조)까지 함께 최적화하는 방식을 주장하지만, 커뮤니티에서는 벤치마크 과최적화에 불과하다는 의심을 받고 있다.
Entity Binding Failures in Tool-Augmented Agents
AI 에이전트가 올바른 도구를 선택해도 잘못된 대상에 실행하는 'Entity Binding 실패' 문제를 정의하고, 이를 막는 실행 정책을 평가한 논문.
Herdr: Agent multiplexer that lives in your terminal
여러 AI 코딩 에이전트(Claude, Codex 등)를 하나의 터미널에서 동시에 실행·관리할 수 있는 Rust 기반 오픈소스 툴로, tmux처럼 세션이 유지되고 SSH로 원격 접속도 가능해 멀티 에이전트 워크플로우를 크게 단순화해준다.
Ornith-1.0: Self-scaffolding LLMs for agentic coding
모델이 문제 풀이 전략(scaffold)을 직접 생성하고 개선하는 자기강화 학습 프레임워크를 적용한 오픈소스 코딩 특화 LLM으로, 9B 소형 모델부터 397B 대형 모델까지 라인업을 갖추고 SWE-Bench 등 주요 벤치마크에서 Claude Opus 4.7을 능가하는 성능을 보여줬다.
Related Resources
- Original blog post
- lecroy-mcp: MCP server for LeCroy oscilloscopes (GitHub)
- spicelib-mcp: SPICE simulator wrapper MCP server (GitHub)
- rc-filter-demo-files: Demo setup files (GitHub)
- kicad-happy: KiCad schematic review automation (mentioned in comments)
- SPICE Wikipedia
- FPAA related EE Times podcast (mentioned in comments)