code-feedback

Code Feedback MCP Server

A powerful, open-source Model Context Protocol (MCP) server for automated, multi-language code feedback, validation, and project analysis. Designed for developers, teams, and LLM-powered workflows to ensure code quality, correctness, and maintainability across diverse languages and build systems.

Why Code Feedback MCP?

Modern codebases are polyglot and complex. Automated, language-agnostic feedback is essential for:

• Catching errors early (before code review or CI)
• Standardizing quality across languages and teams
• Empowering LLMs and tools to validate, lint, build, and analyze code safely
• Reducing manual review time and improving developer confidence

Code Feedback MCP provides a secure, extensible, and LLM-friendly API for code validation, linting, building, and analysis—out of the box.

Key Features

• Multi-language file validation (TypeScript, JavaScript, Python, Go)
• Project-level build/test integration (npm, Make)
• Dependency management for npm projects
• Git command execution
• Secure, path-restricted file and command access
• Structured, machine-readable JSON responses
• Advanced prompt system for code review, analysis, and more
• Cross-platform: Windows, macOS, Linux

How It Works

The MCP server exposes a set of tools and prompts via a simple API. You can:

• Validate and lint code files
• Build and test projects
• Manage dependencies
• Run git commands
• Request advanced code review, analysis, and suggestions via prompts

All actions are path-restricted for security and return structured results for easy integration with LLMs, editors, or CI systems.

Installation

Requirements:

• Node.js 18+
• Language-specific tools (TypeScript, Python, Go, etc.) installed and in your PATH
• Project-level configs (e.g., tsconfig.json, Makefile, package.json) as needed

Install dependencies and build:

npm install
npm run build

Configuration

Configure allowed paths and server settings in mcp-config.json or claude_desktop_config.json:

{
  "mcpServers": {
    "code-feedback": {
      "command": "npx",
      "args": ["-y", "code-feedback"],
      "env": {
        "MCP_ALLOWED_PATHS": "C:/tests"
      }
    }
  }
}

• MCP_ALLOWED_PATHS restricts file/command access for security.

Usage

Start the Server

npm start

Or with the MCP Inspector:

npx @modelcontextprotocol/inspector ./dist/start-server.js

Example: Validate a TypeScript File

Send a request to the server (via HTTP, CLI, or SDK):

{
  "tool": "validate_typescript_file",
  "args": { "filePath": "./examples/example-good.ts" }
}

Response:

{
  "success": true,
  "errors": [],
  "warnings": [],
  "output": "...TypeScript compiler output..."
}

Example: Run an npm Script

{
  "tool": "run_npm_script",
  "args": { "projectPath": "./examples", "scriptName": "test" }
}

Example: Use a Prompt for Code Review

{
  "prompt": "code-review",
  "args": { "filePath": "./examples/example-good.js", "focus": "performance" }
}

Response:

{
  "feedback": "...comprehensive review, suggestions, and best practices..."
}

Available Tools

• validate_typescript_file: Validate and compile a TypeScript file, checking for syntax and type errors.
• validate_javascript_file: Validate JavaScript file syntax using Node.js.
• validate_python_file: Validate Python file with syntax checking and optional linting (pylint, flake8, black, mypy).
• validate_go_file: Validate Go source file with compilation and formatting checks, and optionally run Go tests.
• run_make_command: Run Make commands (e.g., make, make build, make test).
• list_make_commands: List available make targets/commands from a Makefile.
• run_npm_script: Run any npm script defined in package.json (e.g., test, lint, build).
• list_npm_scripts: List all available npm scripts in a project.
• install_npm_deps: Install npm dependencies (packages) in a project.
• uninstall_npm_deps: Uninstall npm dependencies from a project.
• check_npm_dependency: Check if a specific npm dependency is installed in a project.
• run_git_command: Run git commands (status, diff, log, branch, checkout, commit, add, push, pull, merge, reset, or custom).
• uv_init: Initialize a new Python project using uv.
• uv_add: Add Python dependencies to a project using uv.
• uv_run: Run a command in the uv environment.
• uv_lock: Lock Python dependencies using uv.
• uv_sync: Sync Python dependencies using uv.
• uv_venv: Manage the uv virtual environment.
• http: Make HTTP requests (GET, POST, etc.) to localhost or local IPs and return the response.
• docker: Run Docker commands (build, run, stop, rm, rmi, inspect, ps) in a project directory.
• editor: Edit, create, delete, or read text files with robust line/content-based edits, returning git-style diffs.
• filesystem: Secure, batch multi-file/folder CRUD and query operations (delete, create, move, copy, read, stat, search, directory tree, glob support, etc.).
• find: Powerful file and text search using ripgrep (regex, globs, context lines, structured output, etc.).

All tools accept file/project paths and relevant options. Responses are structured as:

{
  "success": true/false,
  "errors": ["error messages"],
  "warnings": ["warning messages"],
  "output": "command output",
  "details": { "command": "...", "exitCode": 0, "duration": 1234 }
}

Usage Examples

uv: Initialize a Python project

{
  "tool": "uv_init",
  "args": { "projectPath": "./my-python-app", "projectName": "my-python-app" }
}

http: Make a local HTTP GET request

{
  "tool": "http",
  "args": { "url": "http://localhost:8000/health" }
}

docker: List running containers

{
  "tool": "docker",
  "args": { "command": "ps" }
}

editor: Replace text in a file

{
  "tool": "editor",
  "args": {
    "action": "edit",
    "file_path": "./README.md",
    "edits": [{ "mode": "content", "oldText": "foo", "newText": "bar" }]
  }
}

filesystem: List directory tree

{
  "tool": "filesystem",
  "args": {
    "ops": [{ "type": "directoryTree", "path": ".", "maxDepth": 2 }]
  }
}

find: Search for TODO comments in JS files

{
  "tool": "find",
  "args": { "pattern": "TODO", "glob": "*.js" }
}

Prompt System

Request advanced code review, analysis, and guidance using prompts:

• code-review: Comprehensive code review with feedback and suggestions
• code-analysis: Analyze code structure, complexity, and potential issues
• refactor-suggestions: Refactoring suggestions for improved code quality
• bug-detection: Detect potential bugs and issues
• documentation-review: Review and suggest improvements for code documentation
• test-coverage-analysis: Analyze test coverage and suggest missing test cases
• security-audit: Security audit and vulnerability identification
• performance-analysis: Performance analysis and optimization suggestions
• coding-task-with-mcp: Guide for completing a coding task using MCP tools and best practices
• find-imports: Find all files that import a specific package
• commit-changes: Show git diff and commit all staged changes with a generated message
• install-and-audit: Install a new npm dependency and check for vulnerabilities
• format-python-directory: Format all Python files in a directory using Black
• build-and-test: Build and test the project using Make or npm
• list-docker-containers: List all running Docker containers
• validate-typescript-file: Validate a TypeScript file for syntax and type errors

Each prompt returns actionable, structured feedback for your code.

When to Use Code Feedback MCP

• After generating or editing code (especially LLM-generated code)
• Before code review or CI to catch issues early
• To automate code quality checks in multi-language projects
• For continuous feedback in editor integrations or bots
• To standardize code review and analysis across teams

Best Practices

• Always validate code with MCP tools after editing or generating code
• Use build and test tools to ensure your code works as expected
• Use prompts for in-depth review, analysis, and improvement suggestions
• Manage dependencies proactively to keep your project clean and up to date
• Reuse existing code and abstractions where possible
• Write modular, generic code to maximize reusability
• Avoid code duplication and hardcoding
• Document the integration and usage of MCP clearly
• Test thoroughly using MCP tools and prompts

Contributing

We welcome contributions! To add a new tool or prompt:

• Fork the repo and create a feature branch
• Add your tool in src/tools/ or prompt in src/prompts/
• Write tests in tests/
• Run npm run lint and npm run test
• Open a pull request with a clear description

Troubleshooting & FAQ

• Ensure all required tools are installed and in your PATH
• Use npm run test to verify your setup
• For permission issues, check script/file permissions and environment
• If you see "Path not allowed", update MCP_ALLOWED_PATHS in your config
• For more help, see the code and examples in the repo

License

MIT

Example

User:

validate the code in D:\Projects\tests\code-feedback-mcp\package.json that it is compile, tests passing, and git status is clean, use code feedback tools

AI:

I'll help you validate the code in your project directory. Let me start by checking what's available and then run through the validation steps.Great! I can access your project directory. Let me first examine the project structure and then run the validation checks.Now let me check the available npm scripts and run the validation steps:Let me start by checking the git status:I can see there are uncommitted changes to package.json. Now let me run the build process to check if the code compiles:Great! The build succeeded. Now let me run the tests:Excellent! All tests are passing. Now let me run the linter to check code quality:Perfect! The linting passed with no issues. Now let me check the current git status again to see the changes:Let me check what changes were made to package.json:## Validation Summary

Here are the results of validating your code-feedback-mcp project:

✅ Compilation Status: SUCCESS

• TypeScript compilation completed successfully
• The build process ran without errors
• Post-build version bump executed correctly (version updated from 0.0.5 to 0.0.7)

✅ Test Status: ALL PASSING

• 19 tests passed across 6 test files
• Test files covered: Go, Python, JavaScript, TypeScript, NPM, and Make tools
• Total test duration: 805ms
• No test failures or errors

✅ Code Quality: CLEAN

• ESLint ran successfully with no issues
• All TypeScript files passed linting with --quiet flag
• No code quality warnings or errors

⚠️ Git Status: NOT CLEAN

• There is 1 uncommitted change: package.json
• The change is a version bump from 0.0.5 to 0.0.7 (automated by the post-build script)
• This appears to be an expected change from the build process

Recommendations:

• Commit the version change: The package.json version bump is from your build process and should be committed
• Consider git workflow: You might want to commit this change to have a clean working directory

If you'd like me to help commit this change or investigate any other aspects of the codebase, just let me know!