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(9120)

0
0
Medium Cost
schimmmi icon
Sec9

This server provides advanced capabilities for n8n workflow management, including compatibility checking, migration, AI-driven analysis, template recommendation and adaptation, and comprehensive security auditing.

Setup Requirements

  • ⚠️Requires an active n8n instance with API access (URL and Key).
  • ⚠️Optional: GitHub Personal Access Token (`GITHUB_TOKEN`) for robust GitHub template discovery and to avoid API rate limits.
  • ⚠️Relies on an internal SQLite database (`~/.n8n_workflow_builder/template_cache.db`) for template caching, which is managed automatically but assumed to be accessible.
Verified SafeView Analysis
The project demonstrates a strong focus on security, actively implementing features like `SecretDetector` to prevent hardcoded credentials, `AuthenticationAuditor` for webhook/HTTP node security, `ExposureAnalyzer` for PII/data exposure, and `RBACManager` for access control and auditing. The `TemplateAdapter` even includes security hardening steps for adapted templates. While the nature of executing 'Code' nodes within n8n workflows inherently carries risk (which the project attempts to mitigate through best practices and detection), the server itself is designed with robust security measures in mind. No obvious malicious patterns or obfuscation were found.
Updated: 2026-01-09GitHub→
0
0
Medium Cost
wingsky-1 icon

email-mcp-server

by wingsky-1

Sec8

Provides email sending, validation, and configuration checking capabilities to AI assistants via the Model Context Protocol.

Setup Requirements

  • ⚠️Requires Python 3.14+.
  • ⚠️Requires an authorization code (for QQ Mail) or an app-specific password (for Gmail), not the regular email password.
  • ⚠️SMTP service must be explicitly enabled in the email provider's settings.
Verified SafeView Analysis
Secrets are handled via environment variables, not hardcoded. Extensive input validation using Pydantic models for email formats, paths, and attachment sizes. Includes an interactive email confirmation feature to prevent unintended sends. Handles local file access (requiring absolute paths and validating them) and remote URL downloads (with size limits and retries), which are inherent risks requiring careful server execution environment and input sanitization.
Updated: 2025-11-23GitHub→
0
0
Medium Cost
Sreenav14 icon

Slack_mcp_server

by Sreenav14

Sec8

Enables AI clients to interact with Slack by providing an MCP server that offers tools for listing channels, sending messages, and fetching message history.

Setup Requirements

  • ⚠️Requires Python 3.x and dependencies like FastAPI, uvicorn, httpx, sqlalchemy, python-dotenv, passlib, pyjwt, and the mcp-sdk.
  • ⚠️Requires creating and configuring a Slack application with appropriate OAuth scopes (chat:write, channels:history, channels:read, groups:read, users:read) to obtain client ID, client secret, and a bot token.
  • ⚠️Requires a PostgreSQL-compatible database connection string for `DATABASE_URL`.
  • ⚠️Users must register an account and then authorize Slack integration through an OAuth flow to use the Slack tools.
Verified SafeView Analysis
Session tokens are passed via URL query parameters for MCP endpoints and the /auth/me endpoint, which is less secure than HTTP headers and risks exposure in logs or browser history. The server lacks explicit rate limiting for authentication and API endpoints, potentially allowing brute-force attacks. Database table creation is handled on application startup (`Base.metadata.create_all`), which is convenient for development but often considered less robust for production deployments compared to dedicated database migration tools (like Alembic, which is present but not configured for autogeneration in the truncated env.py).
Updated: 2025-12-20GitHub→
0
0
Medium Cost
htzzzzzz icon
Sec8

This project implements a Multi-tool Communication Protocol (MCP) server that provides a weather lookup tool and a client that orchestrates LLM function calling with such MCP servers, allowing an AI to dynamically use external tools.

Setup Requirements

  • ⚠️Requires OpenAI/DeepSeek API Key (Paid) for the client's LLM.
  • ⚠️Requires XinZhi Weather API Key (Free tier available, but registration needed) for the server's weather tool.
  • ⚠️Python 3.12+ is required.
  • ⚠️The client requires a `servers_config.json` file to specify MCP server endpoints.
Verified SafeView Analysis
API keys are correctly loaded from environment variables (OPENAI_API_KEY, XINZHI_WEATHER_API_KEY). The server handles different MCP protocol methods and performs basic input validation for the 'city' parameter. No 'eval', code obfuscation, or obvious malicious patterns were detected. Users should ensure any configured MCP server endpoints in `servers_config.json` are trusted.
Updated: 2025-11-27GitHub→
0
0
Medium Cost
nomad-actual icon
Sec4

A comprehensive collection of Model Context Protocol (MCP) tools for personal automation, intelligent information retrieval (RAG, web search, scraping), and common utilities (weather, time, package tracking).

Setup Requirements

  • ⚠️Requires a local Ollama instance (or compatible LLM API endpoint) running for RAG and semantic search functionality.
  • ⚠️Requires a SearXNG instance for web search capabilities.
  • ⚠️Requires Docker and Docker Compose to easily set up and run the server as intended by the provided `docker-compose.yml`.
  • ⚠️Puppeteer, used for screenshots in package tracking, may require additional dependencies or specific configurations in certain environments (e.g., Docker) to function correctly.
Review RequiredView Analysis
The CORS configuration (`app.use(cors)`) is explicitly commented out, which means the server will not enforce any cross-origin restrictions. This is a critical security risk for a network-accessible API, potentially exposing it to Cross-Site Request Forgery (CSRF) and other client-side attacks from any domain. The `puppeteer` usage for package tracking screenshots on arbitrary URLs, while timed out, introduces a potential attack surface if a malicious URL leads to browser exploits or unexpected resource consumption in the underlying Chromium instance. Direct loading of external content into `JSDOM` for web scraping also carries a risk, though `Readability` helps mitigate parsing malicious content.
Updated: 2025-12-15GitHub→
0
0
Medium Cost
Sec1

An agent designed to process or retrieve weather-related data, likely built using the .NET framework and potentially leveraging an AI model.

Setup Requirements

  • ⚠️Requires .NET SDK
Review RequiredView Analysis
No source code provided for analysis. Cannot assess runtime risks like 'eval', obfuscation, or network vulnerabilities. The security score reflects the complete lack of visibility and inability to verify safety.
Updated: 2025-11-19GitHub→
0
0
Low Cost
rishibhushan icon

jenkins_mcp_server

by rishibhushan

Sec8

AI-enabled Model Context Protocol (MCP) server that exposes Jenkins automation through natural-language commands, designed for seamless integration with AI clients like VS Code MCP and Claude Desktop.

Setup Requirements

  • ⚠️Requires Node.js (v14 or higher) for the npx wrapper.
  • ⚠️Requires a Jenkins API Token for authentication, which users must generate manually within Jenkins.
  • ⚠️Corporate network/VPN environments may require specific proxy or Python path configurations, detailed extensively in troubleshooting guides.
Verified SafeView Analysis
The server leverages a Node.js wrapper to spawn a Python process, forwarding CLI arguments directly. The Python backend includes robust input validation before interacting with the Jenkins API. Authentication relies on configurable API tokens or passwords, which are handled via environment variables, .env files, or VS Code settings, with sensitive data masked in logs. While `verify_ssl=False` is a configurable option, it defaults to `True` for secure connections. Explicit warnings against committing `.env` files and recommendations for API token rotation are present in the documentation. No direct `eval` or `exec` on user-controlled input was found. The `_tool_configure_webhook` function's XML modification logic is noted as a placeholder, reducing immediate XML injection risks.
Updated: 2025-12-25GitHub→
0
0
Low Cost
vinaykumar231 icon

Assignment_mcp_server

by vinaykumar231

Sec7

Provides a secure, authenticated interface for an MCP client (e.g., Claude) to interact with Google BigQuery, allowing execution of SQL queries and retrieval of table metadata.

Setup Requirements

  • ⚠️Requires a Google Cloud Platform project with BigQuery API enabled.
  • ⚠️Requires Google Cloud service account credentials (JSON key file) with appropriate BigQuery permissions.
  • ⚠️A strong `SECRET_KEY` must be manually generated and set in the `.env` file.
  • ⚠️Clients must be manually registered using `scripts/register_client.py` to obtain client IDs and secrets and define allowed datasets.
Verified SafeView Analysis
The authentication mechanism uses robust methods (PBKDF2 for secret hashing, JWT tokens for sessions) and access control is implemented at the dataset level. However, the `bq.run_query` tool directly executes arbitrary SQL queries provided by the client. While this is the intended function, it poses a significant SQL injection risk if the calling MCP client is compromised or if the `query` argument is not properly validated/sanitized by the client before being sent to the server. The server itself does not appear to perform SQL sanitization, relying on the client's integrity. No `eval` or `exec` found. Hardcoded secrets are explicitly warned against and the `register_client.py` script helps generate strong credentials.
Updated: 2025-12-06GitHub→
0
0
Medium Cost
ishumilin icon

schwaizer-bfs-mcp

by ishumilin

Sec8

Accessing, searching, and retrieving statistical data from the Swiss Federal Statistical Office (BFS) through various APIs.

Setup Requirements

  • ⚠️Requires Node.js 20.0.0 or higher.
  • ⚠️Initial setup requires running `npm install` to download dependencies.
  • ⚠️Rate limiting from the BFS PXWEB API can be encountered; consider setting `BFS_REQUEST_DELAY` or filtering queries.
Verified SafeView Analysis
The server uses Zod for robust input validation for all MCP tools, which significantly mitigates risks of malformed input or injection attacks. API base URLs are configurable via environment variables, preventing hardcoded secrets for external services. API clients (ky and fetch) include retry logic and timeouts for resilience. Logging is structured and directed to STDERR, keeping STDOUT clean for MCP communication. There are no obvious signs of 'eval', code obfuscation, or malicious patterns. Security relies on the integrity of the underlying BFS APIs.
Updated: 2025-11-20GitHub→
0
0
Medium Cost
Sec7

Runs a local MCP server that provides Google Ads API tools for LLMs to retrieve and analyze advertising account data.

Setup Requirements

  • ⚠️Requires an approved Google Ads Developer Token and an active Google Ads account.
  • ⚠️Requires a Google Cloud Project with the Google Ads API enabled and correctly configured OAuth credentials (Application Default Credentials or Google Ads API Python client library).
  • ⚠️Requires Gemini CLI or Gemini Code Assist for interaction with the MCP server.
Verified SafeView Analysis
The server explicitly states it will expose your Google Ads data to the connected Agent or LLM. It requires broad OAuth scopes (https://www.googleapis.com/auth/adwords, https://www.googleapis.com/auth/cloud-platform) which grant significant access to your Google Ads and Google Cloud resources. Users must be aware of and accept the data exposure risk to the LLM.
Updated: 2025-11-17GitHub→
0
0
Low Cost
Ericwyn icon

cal-tool

by Ericwyn

Sec8

A command-line utility for evaluating mathematical expressions, supporting direct calculations and integration as an MCP server.

Setup Requirements

  • ⚠️Requires Go runtime for building the executable.
  • ⚠️Manual installation step (`mv cal ~/.local/bin/`) is needed for easy command-line access via 'cal'.
Verified SafeView Analysis
The core functionality relies on `github.com/dengsgo/math-engine` to parse and execute expressions. While it doesn't use a general-purpose 'eval', the security of the application is dependent on the robustness of the math-engine against potential expression injection or unexpected side effects. No other obvious network risks, hardcoded secrets, or malicious patterns were identified in the provided source.
Updated: 2025-11-26GitHub→
0
0
Medium Cost
ABaiMate icon
Sec3

This application integrates with the CSDN platform to publish articles programmatically, designed to be callable as a tool by an AI agent.

Setup Requirements

  • ⚠️Requires a valid CSDN API Cookie, which is a sensitive credential obtained from CSDN network requests, and the default in `application.yml` must be replaced.
  • ⚠️Requires Java (likely JDK 17+ for Jakarta EE compatibility) to build and run as a Spring Boot application.
  • ⚠️The hardcoded `x-ca-*` headers in `ICSDNService` might require specific CSDN API key registration or be specific to a certain CSDN API version/access.
Review RequiredView Analysis
CRITICAL RISK: The `application.yml` file contains a hardcoded default CSDN API cookie (`uuid_tt_dd=...; UserName=A_ABai; UserToken=...;`), which is a sensitive user session credential. Running this with the default could expose a specific CSDN account. Although it can be overridden by an environment variable, providing a functional default is a significant security vulnerability. Additionally, `x-ca-key`, `x-ca-nonce`, `x-ca-signature`, and `x-ca-signature-headers` are hardcoded in `ICSDNService.java`; if these are dynamic or user-specific, they also pose a risk.
Updated: 2025-12-02GitHub→
PreviousPage 750 of 760Next