initialize_graphLABSCO SUMMARY
What breaks if I change this becomes a lookup rather than a guess.
An assistant reading files one at a time cannot see callers. With the graph loaded it can: the impact tool answers from the actual call structure, and the dependency tool does the same for modules and files. It holds several repositories at once, which matters when the caller lives in another service.
What it is
A call-graph server built on the nuanced library. It initialises a graph for a Python repository, then answers what calls what, what a file depends on, and what a change would touch.
What you get
- A code graph initialised for a repository path
- More than one repository held at once, with a switch between them and a listing of what is loaded
- The call graph for a named function in a named file
- Module and file dependencies across the codebase
- The impact of changing a specific function
- Graph and function summaries available as resources, plus prompts for function analysis, impact analysis and dependency review
Requirements
Python repositories on local disk - the graph is built from a repository path you give it, and analysis covers Python code.
Setup effort
Build from source — git clone https://github.com/MattMorgis/nuanced-mcp.git
USE CASES
01
Working out what a change will break
Name the function you are about to alter and read back what depends on it, before the test suite tells you.
02
Reading an unfamiliar function in context
Load the repository, then pull the call graph around the function so what it calls and what calls it arrive together.
03
Untangling dependencies before a refactor
Find every module and file the code leans on, then check what moving each piece would touch.
04
Following a call across two repositories
Keep both graphs loaded and switch between them, rather than rebuilding one every time the trail crosses a service boundary.
6 tools
Listed in project docs. Your client may show a different list.
switch_repositorylist_repositoriesget_function_call_graphanalyze_dependenciesanalyze_change_impactBEFORE YOU INSTALL
What to know first
Tools you get
6 tools named in the project's documentation.
Requires
- uv on PATH
- Python 3.13+
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
You clone this one yourself.
Server setup
TERMINAL
git clone https://github.com/MattMorgis/nuanced-mcp.git
cd nuanced-mcpMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"nuanced": {
"command": "uv",
"args": [
"--directory",
"<YOUR_PATH_TO_NUANCED_MCP>",
"run",
"nuanced_mcp_server.py"
]
}
}
}
