UVM 1.2 MCP for Claude

An MCP (Model Context Protocol) server that gives Claude source-code-level knowledge of UVM 1.2 — the Universal Verification Methodology.

When active, Claude can browse actual UVM source files, trace class hierarchies, read method implementations, and retrieve curated expert insights — enabling answers far beyond what general documentation provides.

UVM 1.2 source code and the pre-built knowledge base are bundled in this repository. No additional downloads or build steps required.

What Claude can do with this MCP

| Tool | What it does | |------|-------------| | search_uvm_source | Regex search across all .sv / .svh files | | get_uvm_class | Full class source code | | get_uvm_class_hierarchy | Ancestor / descendant tree | | list_uvm_class_methods | All methods in a class | | get_uvm_method_source | Exact method implementation | | list_uvm_classes | Browse all 100+ UVM classes | | get_uvm_file | Raw file content with line numbers | | list_uvm_files | File tree by subsystem | | get_uvm_macro | Macro definition from uvm_macros.svh | | semantic_search_uvm | Natural-language search (RAG) | | explain_uvm_concept | Deep dive combining source + expert knowledge |

Requirements

Python 3.10+
~200 MB free disk space (for the Python virtual environment)
No GPU required
No external API keys

Installation

1. Clone the repository

git clone https://github.com/YOUR_USERNAME/UVM_MCP.git
cd UVM_MCP

2. Create a virtual environment and install dependencies

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

The first run will automatically download the all-MiniLM-L6-v2 embedding model (~90 MB).

Configure Claude Code

Add the following to ~/.claude/settings.json (create the file if it doesn't exist):

{
  "mcpServers": {
    "uvm": {
      "command": "/absolute/path/to/UVM_MCP/.venv/bin/python",
      "args": ["/absolute/path/to/UVM_MCP/server.py"]
    }
  }
}

Replace /absolute/path/to/UVM_MCP with the actual path where you cloned the repo.

Example (if cloned to your home directory):

{
  "mcpServers": {
    "uvm": {
      "command": "/Users/yourname/UVM_MCP/.venv/bin/python",
      "args": ["/Users/yourname/UVM_MCP/server.py"]
    }
  }
}

Restart Claude Code after saving.

Verify

After restarting, open a new conversation and run:

/mcp

You should see uvm listed as a connected server.

Then try asking Claude:

"Explain how uvm_driver communicates with the sequencer, and show me the relevant source code."

Claude will automatically call the UVM MCP tools to answer with actual source-level detail.

UVM 1.2 Source

Source: accellera/uvm @ UVM_1_2_RELEASE
License: Apache 2.0

MCP Servers