Provides a Git-backed local-remote workspace synchronization engine (WorkBoard) that tracks files, computes unified diffs, pulls and pushes changes, and maintains local Git commit history for RTL and testbench files.
Enables autonomous creation of GitHub issues with full session context, agent model identity, logs, and automatic label normalization via the GitHub API.
Provides low-latency shell execution and atomic file I/O on remote Linux EDA servers over persistent SSH, allowing AI agents to run digital design tool suites and manage files on the remote cluster.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Digital_EDA_MCP run the UVM simulation for my AXI testbench and pull the coverage report"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Model Context Protocol (MCP) Server for Digital IC Design, Formal Property Verification, and Production UVM Verification on Remote Linux EDA Clusters.
Digital_EDA_MCP connects local AI coding assistants (Antigravity, Claude Code, Cursor, Windsurf) to remote Linux EDA computing clusters over persistent, secure SSH and SCP channels. Designed specifically for Digital IC Design, SystemVerilog/UVM Verification, Formal Property Verification, and Static Analysis, it eliminates context bloat and provides lean, purpose-built tools and directives for digital chip design engineers.
┌────────────────────────────────────────────────────────┐
│ Local AI Agent / Copilot │
│ (Antigravity, Claude Code, Cursor) │
└──────────────────────────┬─────────────────────────────┘
│ JSON-RPC (FastMCP / Stdio)
▼
┌────────────────────────────────────────────────────────┐
│ Digital_EDA_MCP │
├────────────────────┬───────────────────┬───────────────┤
│ remote_control │ workboard │ report_issue │
└─────────┬──────────┴─────────┬─────────┴───────┬───────┘
│ Persistent SSH │ Git + SCP │ GitHub API
▼ ▼ ▼
┌──────────────────┐ ┌──────────────────┐ ┌────────────┐
│ Linux EDA Server │ │ Remote Workspace │ │ GitHub │
│ (irun, xrun, jg, │ │ (~/Desktop/UVM/ │ │ Issues │
│ hal, vcs, imc) │ │ or design dir) │ │ Tracking │
└──────────────────┘ └──────────────────┘ └────────────┘Related MCP server: EDA Tools MCP Server
remote_controlLow-latency shell execution and atomic file I/O on the remote Linux EDA server over a persistent, managed SSH session.
run_command: Runs digital tool suites (irun, xrun, ncsim, imc, jg, hal, vcs, yosys, verilator) with environment auto-sourcing (source /cadence/cshrc), timeout management, and stdout/stderr capture.
read_file: Reads remote files (RTL, logs, testbenches, reports) up to 10MB cleanly.
write_file: Atomically creates and updates remote files.
workboardGit-backed local-remote workspace synchronization and version control engine.
Prevents manual SCP file copying and keeps local and remote repositories in lockstep.
Actions:
initialize: Creates a local WorkBoard workspace backed by a dedicated local Git repository.
add: Downloads remote RTL or testbench files via SCP and registers them for tracking.
export: Uploads local files to the remote cluster and registers Git tracking.
pull: Pulls the latest remote versions of tracked files (e.g., simulation logs, IMC coverage reports).
push: Uploads local edits to the remote cluster and automatically creates a local Git commit.
diff: Computes unified diffs between the local working copy and the remote server copy.
status: Summarizes tracking state (SYNCHRONIZED, MODIFIED_LOCALLY, MODIFIED_REMOTELY, CONFLICT).
history: Retrieves the local Git commit log for files or workspaces.
report_issueAutonomous GitHub issue creation tool with full session context, agent model identity, logs, and automatic label normalization.
uvm-architect Skill & SynthesizerBuilt-in production UVM architecture skill located in .agents/skills/uvm-architect/:
Enforces the 12 Golden Tenets of Production UVM Testbenches (clean driver reset routing, lockstep instruction-identity checks, dual-domain coverage sampling, STORE-at-reset commits).
Includes uvm_generator.py for automatically synthesizing complete, production-grade UVM environments from simple YAML/JSON specifications.
Digital_EDA_MCP/
├── server.py # Root entrypoint shim for MCP clients
├── daemon.py # Local FastAPI background daemon for HTTP automation
├── requirements.txt # Python dependencies (FastMCP, Paramiko, FastAPI)
│
├── src/
│ ├── server.py # FastMCP server definition & tool registration
│ ├── issue_reporter.py # Autonomous GitHub issue reporter
│ │
│ ├── core/ # Core Infrastructure Layer
│ │ ├── ssh_client.py # Persistent CSH subshell, sentinels, interactive streams
│ │ └── scp_client.py # OpenSSH SCP direct transfer engine
│ │
│ └── clients/ # Domain Tool Clients
│ ├── workboard_client.py # Git-backed local-remote workspace synchronizer
│ └── eda_client.py # Digital EDA daemon client Python SDK
│
├── .agents/ # Agent Customizations & Skills
│ ├── mcp_config.json # MCP server configuration
│ └── skills/
│ ├── uvm-architect/ # Production UVM environment generator & verification tenets
│ └── eda-mcp-context-router/ # Pre-task context routing directive
│
├── context/ # Operational Agent Specifications
│ ├── designer/ # Digital design & verification execution specifications
│ │ ├── README.md # Designer fast-agent contract & routing
│ │ ├── digital_simulation_guide.md # Incisive/Xcelium, VCS & IMC coverage merge
│ │ ├── formal_verification_guide.md # JasperGold SVA & unbounded proofs
│ │ ├── uvm_architecture_guide.md # Production UVM tenets & reference models
│ │ ├── static_lint_guide.md # Cadence HAL static lint analysis
│ │ ├── workboard_sync_guide.md # WorkBoard file synchronization guide
│ │ └── mcp_tools_spec.md # Tool argument schemas & modes
│ └── coder/ # Server codebase maintenance specifications
│
├── demo/ # Verified Production Demos
│ └── or_Gate/ # Complete UVM verification suite (100% formal, 100% coverage, 100% mutation)
│
├── config/ # Connection configurations (SSH, SCP)
└── tests/ # Unit & integration test suitesPython 3.10+
OpenSSH client installed (ssh, scp)
git clone https://github.com/SiliCAD/Digital_EDA_MCP.git
cd Digital_EDA_MCP
pip install -r requirements.txtCopy the configuration template and specify your remote EDA host and SSH config:
cp config/config.json.template config/config.jsonEdit config/config.json:
{
"ssh_host": "eda-uni",
"ssh_config_path": "~/.ssh/config",
"env_setup_cmd": "source /cadence/cshrc"
}In your MCP settings (e.g. mcp_config.json for Antigravity, Claude Code, or Cursor):
{
"mcpServers": {
"digital-eda-mcp": {
"command": "python",
"args": [
"/path/to/Digital_EDA_MCP/server.py"
]
}
}
}To share remote sessions across external Python automation scripts or test harnesses:
python daemon.pyThe daemon starts an HTTP REST gateway at http://127.0.0.1:8765.
Execute the unit test suite:
python -m unittest tests/test_workboard.py tests/test_issue_reporter.pyDistributed under the MIT License.
Contributions, issues, and feature requests welcome under the SiliCAD organization.
This server cannot be deployed
Persistent memory and cross-session learning for AI coding assistants (hosted remote MCP).
Persistent cloud workspaces for AI agents: run commands, edit files, use git and a browser.
Cross-agent artifact workspace with provenance across Claude Code, Codex, Cursor, LangGraph.
Persistent Linux computers for AI agents: desktop, signed-in browser, vault, apps and shell.
