For AI/ML practitioners working on hardware design automation, KAIST's TERALab workshop is a useful real-world illustration of the Model Context Protocol (MCP) being used to orchestrate AI agents across existing, non-AI-native engineering software, rather than a claim of a new automated-design breakthrough.
What happened
According to The Korea Times, KAIST's TERALab, the semiconductor packaging and interconnect lab led by professor Kim Joung-ho, will run the "HBM Design and Research Automation Workshop Using the OpenClaw AI Agent" on July 3, 2026, from 8 a.m. to noon. Graduate researchers will demonstrate the lab's in-house OpenClaw AI agent platform across two sessions.
Technical context
The first session (8-10 a.m.) covers agents for chip design and simulation: package design and analysis via KiCad MCP, semiconductor power-delivery-network simulation, eye-diagram dataset construction and equalizer optimization, and electromagnetic simulation automation in Ansys HFSS using PyAEDT. The lab will also show knowledge-base construction with Docling and an internal LLM wiki, collaborative multi-agent chatrooms, and automated patch-antenna design. The second session (10:15-11:45 a.m.) focuses on agents for researchers' daily workflows, including portfolio management, automated stock trading through securities-firm APIs, paper-keyword collection, Google Calendar-based presentation drafting, and Linux server management.
For practitioners
The workshop's throughline is MCP as connective tissue between an LLM-based agent layer and pre-existing, specialized engineering tools (Ansys, KiCad) rather than replacing them. That pattern, wrapping legacy tools instead of rebuilding them, is increasingly common wherever agent frameworks meet domain software with deep existing tooling investment.
What to watch
KAIST said it will post recorded demonstrations to the lab's website after the event; those recordings would let outside practitioners assess how robust the MCP-based tool integrations are beyond a live demo setting. The Korea Times, whose article discloses it was produced with generative-AI assistance and edited by its staff, is currently the only outlet to report on the workshop.
Key Points
- 1KAIST's TERALab will showcase its in-house OpenClaw AI agent platform automating HBM chip design and research tasks on July 3, 2026.
- 2The agents use the Model Context Protocol to orchestrate existing EDA tools like Ansys and KiCad rather than replacing them.
- 3The workshop offers practitioners a concrete case study in wrapping legacy engineering software with MCP-based multi-agent orchestration.
Scoring Rationale
A single-lab, single-source workshop demonstrating applied MCP-based AI agent orchestration for chip design automation; a useful applied case study for hardware and MLOps practitioners, but an internal lab demo rather than a released product, paper, or industry-wide shift.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
