The mcp-sandpipersaas server wraps Redwood EDA’s SandPiper-SaaS compiler, exposing every flag and option of the compiler as individual MCP tools. This lets you construct natural-language prompts to drive the SandPiper-SaaS pipeline and introspect its capabilities by calling tools like sandpiper_compile, among others. You can query available tools, inspect their parameters, and invoke the compiler with precise configurations all through the MCP interface. Use this to convert TL-Verilog (TLV) sources into SystemVerilog/Verilog, control optimization levels, target formats, and toggle advanced features via the corresponding tool parameters. The server is managed by UV, which handles the virtual environment and dependency resolution.

Prerequisites:

• Python 3.8+ installed on your system
• Git installed
• UV toolkit installed (uses Python-based UV) or Python 3 environment with uv

Installation steps:

1. Clone the repository: git clone https://github.com/shariethernet/mcp-sandpipersaas.git

2. Install UV (if you don’t have it): pip install --user uv

   or via pipx

   pipx install uv

3. Install dependencies for the MCP server (from the project root): uv install

   This creates a .venv/ and installs required packages

4. Run the MCP server (example with uv): uv --directory <path_to_this_repo>/mcp-sandpipersaas run mcp-sandpipersaas.py

5. Alternative startup with uvx (if you prefer): uvx run mcp-sandpipersaas.py

6. Verify it’s ready by listing tools or invoking a test command via the MCP client.

Notes and tips:

• The server exposes SandPiper-SaaS options as MCP tools. Use tools/list to see available tools and tools/invoke to run a specific tool like sandpiper_compile.
• If you customize paths or environment, ensure <path_to_this_repo>/mcp-sandpipersaas points to the correct project directory.
• Ensure your Python environment has access to the SandPiper-SaaS dependencies; the uv install step will typically handle these within the virtual environment.
• When integrating with clients (Claude/Cline) or other MCP front-ends, you can copy the provided configuration snippet and adjust the directory path to your setup. The autoApprove list allows certain tools to run without additional confirmation (e.g., sandpiper_compile).
• If you encounter issues, verify that the script name matches the file in your repo (mcp-sandpipersaas.py) and that the directory argument points to the correct location.