EDA MCP Server for Yosys, Icarus, OpenLane

Overview

This project provides an MCP (Model Context Protocol) server that exposes common open-source EDA tools — Yosys, Icarus Verilog, OpenLane, GTKWave and KLayout — as programmatic services. The server is intended to be used by AI agents or other automation layers that need to run synthesis, simulation, or ASIC flows and retrieve files, logs, and artifacts in a structured way. Using MCP lets an LLM or assistant call into real EDA tools as part of a larger developer workflow, without manual shell access.

The server typically runs next to your toolchain (or in a container) and manages invocation, file staging, artifact collection, and simple sandboxing. That makes it useful for automated verification loops, reproducible flow runs, and interactive AI-assisted design work where a model issues commands like “synthesize this Verilog” or “run the OpenLane full flow and return GDS.” The server returns standard outputs (stdout/stderr), produced files (netlists, VCD/GHW, GDS), and structured status so the caller can reason about next steps.

Features

• Exposes Yosys for RTL synthesis and netlist extraction
• Runs Icarus Verilog (iverilog + vvp) for simulation and waveform generation
• Orchestrates OpenLane ASIC flows (synthesize, place & route, LVS/DRC) and provides GDS/artifacts
• Produces GTKWave/gtkwave-compatible waveforms and metadata for browsing
• Launches KLayout or prepares files for layout inspection
• File staging: upload inputs and download generated artifacts
• Structured responses: status, logs, and artifact URLs or bundled archives
• Optional Docker/container-friendly deployment for reproducible environments
• Basic sandboxing and resource limits to isolate tool runs

Installation / Configuration

Two common deployment patterns: local Python environment or Docker. Clone the repo and follow one of the examples below.

Clone the repository:

git clone https://github.com/NellyW8/mcp-EDA.git
cd mcp-EDA

Python virtualenv install (example):

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
# Start the MCP server (replace with actual entrypoint if different)
python -m mcp_eda.server --config config.yaml

Example config.yaml

server:
  host: 0.0.0.0
  port: 9999
tools:
  yosys:
    path: /usr/bin/yosys
    timeout: 120
  iverilog:
    path: /usr/bin/iverilog
  openlane:
    path: /opt/openlane
    run_timeout: 7200
artifacts_dir: /var/lib/mcp-eda/artifacts
allow_remote_uploads: true

Docker Compose example

version: "3.8"
services:
  mcp-eda:
    image: nellyw8/mcp-eda:latest
    container_name: mcp-eda
    ports:
      - "9999:9999"
    volumes:
      - ./artifacts:/var/lib/mcp-eda/artifacts
      - /opt/openlane:/opt/openlane:ro
      - /usr/local/bin/yosys:/usr/bin/yosys:ro
    environment:
      - MCP_CONFIG=/etc/mcp-eda/config.yaml

Start with:

docker compose up -d

Configuration tips

• Point tool paths to the installed binaries or mount tool installations into containers.
• Tune timeouts and memory limits to protect the host.
• Use a dedicated artifacts directory with controlled permissions.

Available Resources