Skip to main content
Glama

clenzer

MCP server that hunts dead code and complexity — then cleanses it.
Add it to any CLI agent and your codebase stays lean on every session.

npm version npm downloads License: ISC TypeScript MCP


What it does

clenzer is a Model Context Protocol (MCP) server that plugs into any MCP-compatible CLI agent (Claude Code, Antigravity, Cursor, etc.) and gives it five new tools:

Tool

Description

register_rules

Writes hygiene rules into AGENTS.md / CLAUDE.md so the agent re-enforces them every session

scan_dead_code

Finds unused imports, unused variables, and exported functions with no cross-file references

scan_complexity

Flags long functions, deep nesting, large files, and duplicate code blocks

cleanse

Safely auto-removes dead imports and side-effect-free variables; flags riskier items for manual review

report

Token-efficient summary of all findings with prioritised action items

Design principles

  • Token-efficient — compact output, no JSON blobs, grouped by file

  • Safe — cleanse never removes code with potential side effects; it skips functions and anything with call expressions in initialisers

  • Non-destructive — dry-run mode available; skipped items are always explained

  • Zero config — works on any TS/JS project with or without tsconfig.json


Related MCP server: ts-language-mcp

Installation

Via npx (no install needed)

npx clenzer

Global install

npm install -g clenzer

Local project install

npm install --save-dev clenzer

Adding to your CLI agent

Claude Code / Antigravity CLI

Add to your mcp_config.json (usually ~/.gemini/antigravity/mcp_config.json or ~/.claude/mcp_config.json):

{
  "mcpServers": {
    "clenzer": {
      "command": "npx",
      "args": ["-y", "clenzer"],
      "env": {}
    }
  }
}

Or if installed globally:

{
  "mcpServers": {
    "clenzer": {
      "command": "clenzer",
      "args": [],
      "env": {}
    }
  }
}

Cursor / other MCP hosts

Add the same block to your MCP host's server config. Refer to your host's documentation for the exact file location.


Usage

Once clenzer is connected to your agent, use natural language or call the tools directly.

1. register_rules   — run once per project to lock in hygiene rules
2. scan_dead_code   — before any significant editing session
3. scan_complexity  — identify hotspots
4. report           — get a prioritised action list
5. cleanse          — auto-remove safe dead code

Tool reference

register_rules

project_root: string   # absolute path to project root

Appends clenzer's hygiene rules to AGENTS.md (or CLAUDE.md if it exists). The agent will re-read this file every session, ensuring the rules are always active.

scan_dead_code

project_root: string       # required
include?: string[]         # glob patterns, default: all TS/JS files
exclude?: string[]         # glob patterns, default: node_modules, dist, tests

Reports unused imports, unused variables, and exported functions with no cross-file references. Results are stored in session state for use by cleanse.

scan_complexity

project_root: string       # required
include?: string[]
exclude?: string[]

Thresholds (all configurable via future config file):

  • Function length > 60 lines → long-function

  • Nesting depth > 4 → deep-nesting

  • File size > 600 lines → large-file

  • Duplicate block ≥ 6 lines → duplicate-block

cleanse

project_root: string    # required
dry_run?: boolean       # default: false

Auto-removes items safe to delete (unused imports, variables with no side effects). Skips functions, exports, and anything with call expressions in the initialiser — those are flagged for manual review.

report

project_root: string    # required
top_n?: number          # default: 5 — how many issues to surface per category

Returns a compact markdown summary with dead code counts by kind, high-severity complexity hotspots, and recommended next steps.


Hygiene rules enforced

When you run register_rules, the following are appended to your AGENTS.md:

  1. No unused imports — every import must be referenced in the file body

  2. No unused variables — variables must be read, not just declared

  3. Max function length: 60 lines — extract helpers if exceeded

  4. Max nesting depth: 4 — flatten with early returns

  5. Max file size: 600 lines — split large files into modules

  6. No duplicate code blocks — extract shared logic into utilities

  7. Prefix intentionally unused variables with _ — clenzer skips them


Development

git clone https://github.com/Parth3930/clenzer.git
cd clenzer
npm install
npm run build     # compile TypeScript → dist/
npm run dev       # run with tsx (no compile step)

Project structure

src/
├── index.ts      # MCP server, all 5 tools
├── scanner.ts    # AST-based dead code + complexity analysis (ts-morph)
├── cleanser.ts   # Safe removal engine
├── rules.ts      # AGENTS.md / CLAUDE.md rule injection
└── types.ts      # Shared interfaces

Tech stack

  • @modelcontextprotocol/sdk — MCP server transport and tool registration

  • ts-morph — TypeScript AST analysis and safe code modification

  • zod — Runtime schema validation for tool inputs


Keywords

mcp, mcp-server, dead-code, unused-imports, code-cleanup, refactor, typescript, javascript, code-quality, static-analysis, ast, ts-morph, claude-code, cursor, antigravity, model-context-protocol, linter, cleaner, unused-variables, complexity


License

ISC © Parth3930

Available Tools

5 tools
cleanseA

Safely removes dead code found by scan_dead_code. Auto-removes unused imports and side-effect-free variables. Skips anything risky (functions, exports, initializers with side effects). Always run scan_dead_code first.

Inputs: project_root (required), dry_run.

register_rulesA

Appends clenzer hygiene rules to AGENTS.md (or CLAUDE.md) in the project root so the agent enforces them every session. Run this once when adding clenzer to a project.

Inputs: project_root (required).

reportA

Returns a compact, token-efficient summary of all findings from the last scan. Includes dead code count, complexity hotspots, and actionable next steps.

Inputs: project_root (required), top_n.

scan_complexityC

Scans for unnecessary complexity: overly long functions, deep nesting, large files, and duplicate code blocks.

Inputs: project_root (required), include, exclude.

scan_dead_codeA

Scans a TypeScript/JavaScript project for dead code: unused imports, unused variables, and exported functions with no cross-file references. Results are stored in session state for use by cleanse.

Inputs: project_root (required), include, exclude.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.9/5.0

    Scored across 5 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose: scanning for dead code, cleaning dead code, scanning for complexity, reporting, and registering rules. No overlap.

    Naming Consistency5/5

    All tool names use lowercase with underscores and follow a verb_noun pattern or single verb (cleanse, report). Consistent and predictable.

    Tool Count5/5

    Five tools is well-scoped for a code hygiene assistant, covering scanning, cleaning, reporting, and setup without excess.

    Completeness4/5

    Covers dead code detection and removal, complexity scanning, and reporting. Minor gap: no automatic fix for complexity issues (only scan).

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      B
      maintenance
      Enables AI coding agents to interact with TypeScript projects through compiler-level code intelligence, providing tools for navigation, type information, diagnostics, refactoring, and semantic search.
      29
      194 npm
      4
      Apache 2.0
    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables AI assistants to analyze code health in TypeScript/JavaScript projects, providing tools to run analysis, start a dashboard, and get summaries.
      8 npm
      3
      MIT