Stop Searching. Start Trusting.

The curated directory of MCP servers, vetted for security, efficiency, and quality.

Browse DirectoryWhy Us?

Tired of the MCP "Marketplace" Chaos?

We built MCPScout.ai to solve the ecosystems biggest pain points.

No Insecure Dumps

We manually analyze every server for basic security flaws.

Easy Setup

Our gotcha notes warn you about complex setups.

Avoid "Token Hogs"

We estimate token costs for cost-effective agents.

Products, Not Demos

We filter out "Hello World" demos.

CATEGORIES:
SORT:

Vetted Servers(8554)

0
0
Low Cost
feirate icon
Sec1

A WebSocket server for Slidev to enable real-time updates and control of Markdown content (MCP files), serving as a development tool.

Setup Requirements

  • ⚠️Requires Node.js environment
  • ⚠️Default port 8080 might be in use
  • ⚠️Requires write permissions to the working directory
Review RequiredView Analysis
The server allows unauthenticated WebSocket clients to write arbitrary content to arbitrary file paths on the server's file system via the 'save' event. This is a critical arbitrary file write vulnerability, potentially leading to remote code execution or data corruption.
Updated: 2025-12-11GitHub→
0
0
Low Cost
minghuig icon
Sec10

Facilitate playing chess with an LLM using a Model Context Protocol (MCP) server.

Setup Requirements

  • ⚠️Requires Python 3 and dependencies installed via 'pip install -r requirements.txt'.
  • ⚠️Designed for integration with Claude Desktop, requiring manual modification of the 'claude_desktop_config.json' file for proper setup.
Verified SafeView Analysis
The server uses the well-vetted 'python-chess' library for all game logic and move validation. User inputs for moves are handled by `parse_san` or `from_uci` methods, which are designed to parse chess moves and do not allow for arbitrary code execution. There are no 'eval' statements, network risks (beyond standard local I/O for MCP), or hardcoded secrets. It is designed to run locally via stdin/stdout, mitigating common network attack vectors.
Updated: 2025-11-20GitHub→
0
0
Medium Cost
DevSkillsIT icon

Skills-MCP-WHM-Pro

by DevSkillsIT

Sec9

The MCP Server enables AI assistants to securely interact with WHM/cPanel servers, providing comprehensive automation for web hosting operations, specifically targeting Managed Service Providers (MSPs) and hosting companies.

Setup Requirements

  • ⚠️Requires access to a WHM/cPanel server and a WHM API Token with appropriate permissions.
  • ⚠️Node.js 18.0.0 or higher is required.
  • ⚠️Destructive operations (e.g., deleting accounts, resetting DNS zones) require the MCP_SAFETY_TOKEN environment variable to be set and passed as 'confirmationToken' in the request.
Verified SafeView Analysis
The project exhibits robust security measures, including timing-safe API key comparisons, extensive sensitive data sanitization in logs, strict whitelisting for SSH commands and log file access (preventing RCE and unauthorized file access), comprehensive path validation against directory traversal and shell metacharacters, and an explicit 'Safety Guard' system requiring a confirmation token for all destructive operations. It also integrates ACL validation and optimistic locking for DNS. While inherently managing a powerful hosting platform carries risk, the mitigations implemented are excellent.
Updated: 2025-12-20GitHub→
0
0
Low Cost
abehmiel icon
Sec9

A self-hosted Model Context Protocol (MCP) server that extends Claude Desktop's context window by providing Retrieval-Augmented Generation (RAG) capabilities over personal documents and codebases on macOS.

Setup Requirements

  • ⚠️Requires macOS 11.0 or later for operation and GPU acceleration (MPS).
  • ⚠️Requires 'uv' Python package manager (v0.5.24+) to be installed globally.
  • ⚠️Requires Claude Desktop app to be installed for integration.
Verified SafeView Analysis
The project emphasizes local execution with no external API keys, significantly reducing network attack surface and data leakage risks. Input validation is performed for document addition. Subprocess calls in utility scripts are for known, controlled commands (e.g., `uv`, `pytest`) and not exposed to arbitrary user input. Configuration parsing includes JSON syntax validation and backup creation. No 'eval' or direct hardcoded secrets were identified.
Updated: 2025-12-02GitHub→
0
0
Low Cost
QWED-AI icon

qwed-mcp

by QWED-AI

Sec3

Provides deterministic verification tools for LLM outputs to enhance trustworthiness in mathematical calculations, logical reasoning, code safety, and SQL queries.

Setup Requirements

  • ⚠️Requires SymPy and Z3-solver Python libraries.
  • ⚠️Requires Python 3.10+.
  • ⚠️Manual configuration of MCP client (e.g., Claude Desktop JSON file) is necessary for integration.
Review RequiredView Analysis
The `verify_math_expression` function, which accepts mathematical expressions and claimed results directly from LLMs, utilizes SymPy's `parse_expr`. `parse_expr` is a known vector for remote code execution if supplied with malicious input, as it can be used to execute arbitrary Python code (e.g., via `__import__`). Although a `local_dict` is passed, it does not fully mitigate this risk, making the server vulnerable when processing untrusted mathematical expressions. The `verify_code` and `verify_sql` engines are designed for security detection, which is positive, but the math engine poses a critical vulnerability.
Updated: 2026-01-18GitHub→
0
0
Low Cost
dmarsters icon
Sec10

Enhances image generation prompts with microscopy aesthetic vocabulary.

Setup Requirements

  • ⚠️Requires Python >=3.9
  • ⚠️Relies on FastMCP framework runtime for tool invocation
  • ⚠️Requires `pip install -e ".[dev]"` for initial setup
Verified SafeView Analysis
The server code handles inputs with validation against predefined lists, preventing injection or unexpected behavior. There are no instances of `eval`, `exec`, or other dynamic code execution from user input. No hardcoded secrets are present. JSON output is properly handled with `json.dumps`. The use of the `fastmcp` framework further suggests a controlled tool invocation environment.
Updated: 2025-11-29GitHub→
0
0
Medium Cost
abel-vs icon

macadamia-mcps

by abel-vs

Sec8

This is a custom Minecraft server implementation providing core gameplay mechanics, administrative commands, and custom features like ranks, homes, warps, and anti-swear filters, aimed at creating a unique server experience.

Setup Requirements

  • ⚠️Requires a MySQL Database
  • ⚠️Requires a compatible Minecraft Server (e.g., Spigot, Paper) to run as a plugin
  • ⚠️Highly recommended: Vault Plugin for full economy functionality
Verified SafeView Analysis
The server uses `PreparedStatement` for database interactions and performs permission checks for administrative commands, which are good security practices. Configuration details, including MySQL credentials, are loaded from a `config.yml` file, mitigating hardcoded secrets in the source code. No 'eval' or obvious obfuscation detected. Potential areas for further review include comprehensive SQL injection defenses across all queries, robustness of chat filters, and exhaustive permission validation for all user-facing commands, though initial checks appear solid.
Updated: 2026-01-19GitHub→
0
0
Low Cost
nbence94 icon

mcp-framework

by nbence94

Sec7

This project enables Large Language Models (LLMs) to perform browser automation for web application testing and interaction by exposing Playwright actions as callable tools.

Setup Requirements

  • ⚠️Requires Python 3.14+ (as specified in `pyproject.toml`. Python 3.14 is an unreleased version, suggesting a potential typo or future-proofing, which may prevent the project from running on current Python installations).
  • ⚠️Requires `playwright install` to download browser binaries after `pip install playwright`.
  • ⚠️Specific configuration (`claude_desktop_config.json`) is needed for integration with Claude Desktop.
Verified SafeView Analysis
The server stores demo application credentials (usernames and passwords for Saucedemo and OrangeHRM) directly in YAML configuration files. While these specific credentials are for public demo sites and are not highly sensitive, this pattern of storing credentials in plaintext configuration files is generally discouraged for production environments with sensitive data. No 'eval' or obvious obfuscation found. The primary security consideration for a browser automation framework controlled by an LLM is the potential for misuse by the LLM itself (e.g., navigating to arbitrary sites or performing unintended actions), which is an inherent risk of such a system rather than a vulnerability in the server's code.
Updated: 2026-01-18GitHub→
0
0
Medium Cost
ManInBlack-coder icon

mcp-servers-setup

by ManInBlack-coder

Sec8

Setting up and managing containerized MCP (Myriad Core Platform) servers for platforms like GitHub and Bitbucket.

Setup Requirements

  • ⚠️Docker is required.
  • ⚠️A `.env` file containing `GITHUB_TOKEN=your_pat_token` is required to run the GitHub MCP server via the provided script.
  • ⚠️Specific platform credentials (e.g., GitHub Personal Access Token, Bitbucket Username/App Password) are required for the respective MCP servers.
Verified SafeView Analysis
No 'eval' or obfuscation found in the provided code snippets. Secrets are handled through environment variables and `.env` files, which is a good security practice. No hardcoded credentials are visible. The security depends on the integrity of the underlying Docker images (`ghcr.io/github/github-mcp-server:latest`, `bitbucket-mcp-server:latest`).
Updated: 2025-11-25GitHub→
0
0
Low Cost
miandin icon
Sec9

A Node.js server that authenticates users with predefined credentials, issues JWTs, and acts as a secure proxy to a QRadar MCP endpoint by acquiring and attaching an OAuth token to forwarded requests.

Setup Requirements

  • ⚠️Requires Node.js environment to run.
  • ⚠️Crucial environment variables (`MCP_SERVER_API_KEY_USERNAME`, `MCP_SERVER_API_KEY_PASSWORD`, `JWT_SECRET`, `PROXY_TARGET_URL`, `API_CLIENT_ID`, `API_CLIENT_SECRET`, `API_TOKEN_URL`) must be correctly configured, preferably in a .env file or system environment.
Verified SafeView Analysis
No 'eval' or obfuscation found in the provided source code. All sensitive configurations (credentials, secrets, target URLs) are correctly sourced from environment variables using `dotenv`. The authentication flow uses standard JWTs. Security relies on strong, securely managed environment variables for secrets like `JWT_SECRET`, `MCP_SERVER_API_KEY_USERNAME`, and `MCP_SERVER_API_KEY_PASSWORD`. There are no apparent hardcoded secrets within the code itself, which is good practice.
Updated: 2025-11-24GitHub→
0
0
Medium Cost
agorski icon
Sec9

A template for developing custom Model Context Protocol (MCP) servers in JavaScript for integration with Claude Desktop, providing tools for server information, basic greetings, weather lookups, and historical stock data.

Setup Requirements

  • ⚠️Node.js (v16 or later) is required.
  • ⚠️Requires manual configuration in Claude Desktop's `claude_desktop_config.json` file.
  • ⚠️Relies on external public APIs (Open-Meteo and Yahoo Finance) for weather and stock data, which could experience downtime or changes.
Verified SafeView Analysis
The server code does not use 'eval' or any obfuscation. It performs external network requests for its 'weather' and 'stock' tools to public APIs (Open-Meteo and Yahoo Finance), which is expected functionality for these tools. There are no apparent hardcoded secrets or malicious patterns. Basic error handling is implemented for tool arguments and external API failures. The server is designed for local execution and communication via standard I/O.
Updated: 2025-12-14GitHub→
0
0
Medium Cost
nikosolihin icon
Sec8

A remote MCP server that extracts transcripts, titles, descriptions, and comments from Loom videos to augment AI agent capabilities.

Setup Requirements

  • ⚠️Requires a Cloudflare Workers account for deployment.
  • ⚠️Requires Cloudflare Wrangler CLI configured.
Verified SafeView Analysis
The server is explicitly 'authless', meaning it does not implement its own authentication or authorization. Anyone with the deployed server's URL can query it. While this is intentional for accessing public Loom video data, it means there are no access controls on who can trigger these Loom API calls. No hardcoded secrets, 'eval', or obvious malicious patterns are present in the source code. Network requests are limited to the legitimate Loom GraphQL API.
Updated: 2026-01-19GitHub→
PreviousPage 560 of 713Next