pyscn MCP Server: Setup for Claude Code, Cursor and Other Apps

• Runs every code-quality check on a Python file or folder

  Runs several analyses in one pass: complexity, dead code, duplicate code, class coupling and cohesion, import dependencies and code communities. You pick which ones and whether you want a short summary or the full report.

  Try: “Run a full pyscn analysis on ./src and summarize the worst problems.”

  Files · analyze_code

• Finds Python functions that are too complex to maintain

  Measures cyclomatic complexity for each function and class scope and lists the ones above a limit you set (default 10).

  Try: “Which functions in ./app have a complexity above 15?”

  Files · check_complexity

• Finds copy-pasted or near-identical Python code

  Compares code structure to spot duplicated blocks, even when names differ. You set the similarity cutoff (default 0.8) and the smallest block size to report.

  Try: “Find duplicated code in ./src that is at least 10 lines long.”

  Files · detect_clones

• Shows classes that depend on too many other classes

  Counts how many other classes each class depends on (CBO) and reports the ones over your threshold (default 10).

  Try: “Which classes in ./src are too tightly coupled to others?”

  Files · check_coupling

• Shows classes that do too many unrelated things

  Scores how well each class's methods share state (LCOM4). A high score suggests the class should be split.

  Try: “List classes in ./src with poor cohesion.”

  Files · check_cohesion

• Finds code that can never run

  Builds a control-flow graph for each function and reports unreachable statements, filtered by a minimum severity (default warning).

  Try: “Find unreachable code in ./src.”

  Files · find_dead_code

• Gives a 0 to 100 health score with a grade

  Returns one overall score and a grade, plus separate scores for complexity, dead code, duplication, coupling, cohesion, dependencies and architecture.

  Try: “What is the health score of this repo?”

  Files · get_health_score

• Spots dependency-injection mistakes in Python classes

  Looks for constructors that take too many parameters (default limit 5), hidden dependencies, concrete dependencies and service-locator use.

  Try: “Check ./src for dependency injection anti-patterns.”

  Files · detect_di_antipatterns

Watch out for

• Reads every Python file under the path you give

  Each tool takes a file or folder path and analyzes the code in it. Point it at the project, not at your home folder. github.com

• Source code goes to your AI app as results

  Findings include file paths and code locations, which your assistant then sees. Treat them like any code you paste into the chat. github.com

Security advisories