Stop Searching. Start Trusting.

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

Browse Directory Why 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(9120)

Medium Cost

mcp-secure-shell

by Brainwires

Sec8

Enables AI agents to execute shell commands and transfer files on remote hosts via SSH/SFTP, acting as an MCP server.

Setup Requirements

⚠️Requires SSH key-based authentication to be configured for target hosts.
⚠️SFTP must be enabled on remote servers for file transfer (e.g., `Subsystem sftp internal-sftp` in sshd_config).
⚠️Requires Rust toolchain for building/installation.

Verified SafeView Analysis

The server uses standard and robust SSH/SFTP libraries, enforcing key-based authentication and host key verification against ~/.ssh/known_hosts. The primary security consideration is the inherent risk of allowing an AI agent to execute arbitrary commands on remote hosts, which grants it significant power. Users must ensure their SSH configurations, key management, and remote access policies are secure, as the server operates within these established SSH security contexts rather than introducing new vulnerabilities.

Updated: 2025-12-14GitHub→

High Cost

Cyber_MCPs

by hackersatyamrastogi

Sec3

This project provides a collection of Model Context Protocol (MCP) servers that integrate various security tools, enabling AI assistants to perform security assessments through natural language interactions.

Setup Requirements

⚠️Requires various external security tools (e.g., Nmap, Nuclei, SQLmap) to be installed on the system where the MCP server runs.
⚠️Some tools (e.g., SQLmap, Commix, ScoutSuite) require Python to be installed.
⚠️Specific servers require API keys or credentials configured via command-line arguments (e.g., MobSF) or environment variables (e.g., Nessus if not in mock mode).
⚠️The `scoutsuite` server is vulnerable to arbitrary code execution due to its method of parsing external tool output.

Review RequiredView Analysis

The server explicitly warns about critical security risks including command injection, privilege escalation, network exposure, and data exfiltration. While most tool wrappers use `child_process.spawn` or `node-pty.spawn` with array arguments (which is generally safer against shell injection than string interpolation) and inputs are validated with Zod, a severe vulnerability exists in the `scoutsuite` server. It uses `vm.runInContext()` to execute JavaScript code read directly from a file generated by the external `ScoutSuite` tool. If this external file (`scoutsuite_results_*.js`) were compromised or contained malicious JavaScript, it would lead to arbitrary code execution within the MCP server's Node.js process. No obvious hardcoded secrets were found, but some tools require API keys/credentials as arguments or environment variables. The inherent nature of these security tools involves network interaction and can require elevated privileges, as extensively detailed in the project's `SECURITY.md`.

Updated: 2025-12-05GitHub→

High Cost

uc-mcp

by 33jrdd

Sec8

Programmatically interact with the UnknownCheats forum, bypass Cloudflare protection, and extract structured data including thread content and code blocks.

Setup Requirements

⚠️Requires Google Chrome installed locally.
⚠️Requires Node.js 18 or higher.
⚠️Opens a visible Chrome browser window (due to `headless: false`).

Verified SafeView Analysis

The server uses `puppeteer-real-browser` to control a local Chrome instance, which introduces dependency on Chrome's security and the websites it navigates. The `performLogin` function handles sensitive credentials (username/password) provided by the LLM, but these are not hardcoded. Session cookies are saved locally to `cookies.json`, which could be a sensitive file if not properly secured on the host system. No direct use of `eval` or malicious patterns observed in the provided core source code, though `pac-resolver` (a dependency) uses `degenerator` which can dynamically generate functions for PAC file processing.

Updated: 2026-01-19GitHub→

Medium Cost

agrivision-mcp-server

by eagleisbatman

Sec9

AI-powered plant disease and health diagnosis using vision models via Model Context Protocol (MCP).

Setup Requirements

⚠️Requires Google Gemini API Key (Paid service)
⚠️Requires Node.js runtime (>=18.0.0)
⚠️Relies on an external Crops API for dynamic crop lists (with fallback)

Verified SafeView Analysis

The server demonstrates good security practices. It loads API keys from environment variables, performs robust input validation (image format, size, guardrails for content safety), configures CORS, and implements timeouts for external API calls to prevent hanging. Error handling is structured to avoid exposing internal details. There is no evidence of 'eval', obfuscation, or direct command injection vulnerabilities in the provided source code.

Updated: 2025-12-05GitHub→

Low Cost

vonage-mcp-server

by mobilebiz

Sec3

Provides Vonage SMS, CSV bulk SMS, and voice call functionalities as a Model Context Protocol (MCP) server for AI assistants.

Setup Requirements

⚠️Requires Node.js version 22.0.0 or higher.
⚠️Requires manual setup of Vonage Application ID, Private Key file, and Voice Call From Number as environment variables.
⚠️The HTTP API wrapper, if exposed publicly, has a critical authentication vulnerability as it uses the non-secret Vonage Application ID as an API key.

Review RequiredView Analysis

The HTTP wrapper (`src/http-server.ts`), designed for integration with external applications like Dify, uses `process.env.VONAGE_APPLICATION_ID` as the API key for authentication (`X-API-KEY` header). As the `VONAGE_APPLICATION_ID` is an identifier and not a secret, it can be easily discovered or exposed, leading to unauthorized access and control of Vonage services (SMS sending, voice calls) through the server's HTTP endpoints. While private keys are loaded from files and other secrets via environment variables, this weak authentication for HTTP exposure is a critical vulnerability. No signs of 'eval', obfuscation, or other overt malicious patterns were found. For local usage via StdioServerTransport (e.g., with Claude Desktop), the immediate network exposure risk is significantly lower.

Updated: 2025-12-10GitHub→

Medium Cost

rubocop-mcp-server

by ishtiaque05

Sec9

Provides Ruby RuboCop and Rails linting capabilities to AI assistants via the Model Context Protocol (MCP) for code quality and auto-correction.

Setup Requirements

⚠️Requires RuboCop and rubocop-rails gems to be installed via `gem`.
⚠️Requires Node.js version >= 20.17.0.
⚠️Requires Yarn package manager for installation.

Verified SafeView Analysis

The server executes external `rubocop` commands using `child_process.execFile`. Arguments passed to `rubocop` (like file paths, cop names) are constructed from user input. `execFile` is generally safer than `exec` as it prevents shell injection by not interpreting arguments as shell commands. The risk of command injection is low, assuming the `rubocop` binary itself is trusted and the arguments passed are treated as literal strings by `rubocop`. No hardcoded secrets or direct network listen ports are identified, as it primarily communicates via standard I/O.

Updated: 2025-12-18GitHub→

Low Cost

test-remote-mcp-server

by arunkumar-maker

Sec9

This server provides a simple calculator API for basic arithmetic operations and random number generation.

Setup Requirements

⚠️Python 3.12+ only
⚠️Requires fastmcp library

Verified SafeView Analysis

The server's source code is simple and performs basic arithmetic and random number generation. There are no obvious signs of 'eval' or similar dangerous functions, obfuscation, or hardcoded secrets. The 'os' module is imported but not utilized, minimizing potential system-level risks. The server binds to 0.0.0.0, which is standard for accessibility but implies it will be reachable from external networks if deployed without proper firewalling. Given its limited functionality, the attack surface is minimal.

Updated: 2026-01-19GitHub→

Low Cost

mcp-server

by seeeeeeong

Sec9

A template for building custom Model Context Protocol (MCP) servers to extend AI model capabilities.

Setup Requirements

⚠️Requires Python 3.12 or higher
⚠️Requires `uv` package manager

Verified SafeView Analysis

The provided source code is a template and implements a very simple example tool. It does not contain 'eval' or similar dangerous functions. Communication is handled via standard I/O (stdio_server), limiting direct network exposure. The `handle_call_tool` processes arguments from client input, but in a controlled manner (dictionary lookup, string formatting) without obvious code injection vulnerabilities. Customization with new tools or resources could introduce security risks if not carefully implemented, but the current code is secure.

Updated: 2025-12-02GitHub→

Low Cost

mcp-smart-reader

by prefrontal-systems

Sec8

Enables token-efficient interaction with large documents for LLM agents by providing automatic summarization and granular section extraction.

Setup Requirements

⚠️Requires Python 3.10+
⚠️Requires fastmcp library
⚠️Requires tiktoken library

Verified SafeView Analysis

The server's `smart_read`, `read_section`, and `list_sections` tools accept a `file_path` argument, allowing it to read arbitrary local files on the system where it's running. While this is core functionality for a document reader, it means the server should be run in a trusted, isolated, or sandboxed environment if exposed to untrusted input to prevent unauthorized file access. No 'eval', obfuscation, hardcoded secrets, or directly malicious patterns were found in the provided code.

Updated: 2025-12-06GitHub→

Medium Cost

MCP-server-

by vilashkardate

Sec1

Hosts or manages a Minecraft game server.

Setup Requirements

⚠️Requires Java Runtime Environment (JRE) for execution (typical for Minecraft servers).
⚠️No specific setup instructions provided due to missing source code.

Review RequiredView Analysis

CRITICAL: No source code was provided for analysis. Therefore, a comprehensive security audit could not be performed. The score reflects an inability to verify safety, making it inherently risky to run any unknown code.

Updated: 2025-11-20GitHub→

Low Cost

gmail-mcp-server

by okorie2

Sec7

Provides AI assistants with tools to fetch user Gmail profiles and send emails through API interactions.

Setup Requirements

⚠️Requires a Google Cloud Project with Gmail API enabled.
⚠️OAuth 2.0 Client IDs (GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET) must be configured as environment variables.
⚠️Requires initial user authentication via a web browser, and re-authentication on every server restart as tokens are not persisted.
⚠️Node.js (v18+ recommended).

Verified SafeView Analysis

The server handles sensitive Google OAuth tokens. `GOOGLE_CLIENT_ID` and `GOOGLE_CLIENT_SECRET` are read from environment variables, which is a good practice. OAuth tokens are *not* persisted to disk (the `fs.writeFileSync` call is commented out), which significantly reduces the risk of static token compromise but necessitates re-authentication on every server restart. The requested Gmail API scopes (`gmail.send`, `gmail.readonly`) grant broad access to user emails, which is a high-sensitivity area. Error messages logged to console could expose internal details if the server is improperly secured or monitored.

Updated: 2025-11-23GitHub→

Medium Cost

slovlex-mcp-server

by tomique34

Sec5

This project appears to be a custom server implementation for the Minecraft protocol, likely used for hosting custom game experiences.

Review RequiredView Analysis

Source code was not provided for analysis, preventing a comprehensive security audit. A score of 5 is assigned as a neutral placeholder. Specific vulnerabilities like hardcoded secrets, 'eval' usage, or malicious patterns could not be identified. Running this project without code review is not recommended.

Updated: 2025-12-01GitHub→

PreviousPage 454 of 760Next