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
deeprave icon
Sec9

A minimal MCP server providing a web search tool via DuckDuckGo, designed for integration with AI agents.

Setup Requirements

  • ⚠️Python 3.10+ required
  • ⚠️Requires manual MCP client configuration specifying the absolute path to the 'search_mcp.py' script.
  • ⚠️Relies on DuckDuckGo's public web interface for scraping, which can be less stable or subject to rate limiting/blocking compared to a dedicated API.
Verified SafeView Analysis
The server uses stdio for communication, limiting direct network exposure to external clients. It does not contain 'eval', hardcoded secrets, or obfuscation. The primary external interaction is through the 'ddgs' library which scrapes DuckDuckGo, inherently relying on the stability and structure of DuckDuckGo's website, which could be a point of failure for the functionality, but not a direct security vulnerability within the server's code itself. Error handling for network issues and timeouts is present.
Updated: 2025-12-11GitHub→
0
0
Low Cost
Akungapaul icon

wp-menus-mcp

by Akungapaul

Sec8

This server enables external applications to manage WordPress navigation menus via the Model Context Protocol (MCP).

Setup Requirements

  • ⚠️Requires Node.js version 18 or higher.
  • ⚠️Requires a running WordPress instance with the REST API enabled.
  • ⚠️Requires a WordPress username and an application password with sufficient permissions to manage navigation menus and potentially site settings.
  • ⚠️Integration with the Model Context Protocol (MCP) ecosystem is implied by the SDK dependency.
Verified SafeView Analysis
The server utilizes environment variables for sensitive WordPress credentials (URL, username, app password), which is good practice for managing secrets. Input validation is performed using `zod` schemas for each tool, which helps mitigate common injection vulnerabilities. The `set_menu_location` tool directly interacts with a sensitive WordPress settings endpoint (`/wp/v2/settings`). While the input is validated, misconfiguration of WordPress permissions for the provided application password or insufficient security measures for the MCP server itself could lead to unauthorized menu changes or broader site settings modifications. No 'eval' or obvious obfuscation is present in the provided code.
Updated: 2025-11-28GitHub→
0
0
Low Cost
baptiste-max icon

holded-mcp-server

by baptiste-max

Sec1

Source code not provided, cannot determine the specific use case of this 'holded-mcp-server'.

Setup Requirements

  • ⚠️Source code not provided, cannot identify specific requirements or friction points for setup.
Review RequiredView Analysis
Critical: Source code was not provided. Therefore, a security audit cannot be performed. It is impossible to assess risks such as 'eval' usage, obfuscation, network vulnerabilities, hardcoded secrets, or malicious patterns without the code itself. Running any software without its source code being audited is highly risky.
Updated: 2025-12-06GitHub→
0
0
Medium Cost
Harikapudotam icon

mcp-server-cap-odata

by Harikapudotam

Sec6

An MCP server for CAP OData services that translates natural language requests into JSON-RPC calls for OData CRUD operations using Google Gemini, acting as an AI gateway to enterprise data.

Setup Requirements

  • ⚠️Requires a `GEMINI_API_KEY` environment variable for AI functionality, which is a paid service.
  • ⚠️The server explicitly disables SSL certificate validation (`rejectUnauthorized: false`) for outgoing HTTPS requests, which is a critical security concern for production deployments.
  • ⚠️The project is built on SAP CAP (Cloud Application Programming Model), implying knowledge of the SAP ecosystem for full utilization and deployment.
Review RequiredView Analysis
1. **Critical Vulnerability (`rejectUnauthorized: false`):** The `https.Agent` is configured with `rejectUnauthorized: false`. This disables SSL certificate validation for all outgoing HTTPS requests made via `axios`. In a production environment, this is a severe security risk, making the application vulnerable to Man-in-the-Middle (MITM) attacks where an attacker could intercept and potentially alter communication with the backend OData service. 2. **Functional Mismatch in OData Query Construction:** There is a discrepancy between the structured filter and sort parameters defined in `srv/rpcSchema.js` and the format expected by `srv/tools/read.js` and `srv/tools/query.js`. The tool functions appear to expect OData query string components (e.g., `filter` as a string), but `normalizeToJsonRpc` and the `RPC_SCHEMA` define these as structured objects. This functional bug will likely lead to runtime errors or incorrect query results for complex read/query operations. 3. **Good Input Validation:** The server uses `Ajv` and a well-defined `RPC_SCHEMA` for validating incoming JSON-RPC requests, and Google Gemini's structured output feature is leveraged to ensure natural language inputs are transformed into schema-compliant JSON-RPC. This is a good practice for mitigating injection risks from malformed requests.
Updated: 2025-12-08GitHub→
0
0
Medium Cost
RazonIn4K icon

grok-mcp-server

by RazonIn4K

Sec9

Provides AI assistants with access to Grok 4's capabilities including chat completions, live search, and model management via the Model Context Protocol.

Setup Requirements

  • ⚠️Requires an xAI API Key (paid service)
  • ⚠️Node.js >= 18.0.0 is a prerequisite
  • ⚠️Optional setup with 'direnv' for environment variable management adds a dependency and configuration step
Verified SafeView Analysis
The server demonstrates strong security practices including explicit use of environment variables for API keys, robust input validation via Zod, and a basic sanitization function to mitigate XSS/prototype pollution. Communication occurs over standard I/O (stdio) as per the MCP specification, avoiding direct network exposure for tool calls. Resilience features like rate limiting and circuit breakers are implemented. The shared secret authentication mechanism is present but currently commented out in the provided code.
Updated: 2025-11-24GitHub→
0
0
Medium Cost
happyvertical icon

smrt-docs-mcp

by happyvertical

Sec9

Provides AI-powered access to SMRT framework documentation, routing developer queries to relevant package experts using RAG (Retrieval-Augmented Generation) from CLAUDE.md files.

Setup Requirements

  • ⚠️Requires an AI provider API key (e.g., OpenAI, Anthropic, Gemini) which may incur costs.
  • ⚠️Requires Node.js runtime version 24.0.0 or higher.
  • ⚠️Requires the SMRT framework source code to be present locally for scanning CLAUDE.md documentation files.
Verified SafeView Analysis
The server uses environment variables for AI API keys, preventing hardcoded secrets. Communication with the MCP client is via standard I/O (stdio), which is generally secure within a controlled environment. It reads local CLAUDE.md files for documentation, which is within its expected operational scope. No 'eval' or other direct code execution vulnerabilities were found within the provided source. The 'auto-changeset.ts' script executes git commands, but this is a development/CI script and not part of the runtime server.
Updated: 2025-12-10GitHub→
0
0
Low Cost
RolleRocker icon

mcp-demo-server

by RolleRocker

Sec7

Demonstrates the Model Context Protocol (MCP) capabilities by offering tools for arithmetic, note management, weather queries, and file system operations, along with dynamic resources and prompt templates.

Setup Requirements

  • ⚠️Requires Java 21 or higher
  • ⚠️The `read_file`, `write_file`, and `list_directory` tools interact with the local file system where the server is run, subject to internal path validations.
Verified SafeView Analysis
The server implements file system operations (`read_file`, `write_file`, `list_directory`). Path traversal attempts ('..') and access to known sensitive system directories (`/etc`, `/sys`, `/proc`) are explicitly mitigated by the `FilePathValidator` in the domain layer. No hardcoded secrets or direct remote code execution vulnerabilities were identified. The external weather API (Open-Meteo) is free and does not require an API key. File I/O operations inherently carry risk, but the explicit validations enhance safety.
Updated: 2025-12-27GitHub→
0
0
Medium Cost
CuteDandelion icon

VPS-Execute-MCP

by CuteDandelion

Sec8

Enables AI agents and other clients to securely interact with remote servers via SSH for command execution, file management, and system monitoring.

Setup Requirements

  • ⚠️Requires SSH access to a remote host, configured with either a private key (recommended) or password.
  • ⚠️Requires Node.js 18+ to run.
  • ⚠️The `.env` file must be correctly configured with SSH credentials (host, username, and authentication method) and API_KEY (for REST API) before running. For Claude Desktop integration, an absolute path to the `mcp-server.js` file and its environment variables must be configured in Claude's settings.
  • ⚠️SSH private key passphrases, if used, will be stored in plain text in the `.env` file or Claude Desktop configuration, which is a security consideration.
Verified SafeView Analysis
The server implements several good security practices, including API key authentication, rate limiting, and comprehensive audit logging. It supports SSH key authentication (recommended) and includes basic dangerous command pattern detection, with an optional command whitelisting feature. However, storing SSH private key passphrases in plain text environment variables is a potential risk, though acknowledged in documentation. The overall security heavily relies on proper user configuration (e.g., using a dedicated SSH user with limited permissions and enabling command whitelisting).
Updated: 2025-12-13GitHub→
0
0
Medium Cost
filiperochalopes icon

esus-pec-mcp

by filiperochalopes

Sec9

Serves as a Model Context Protocol (MCP) endpoint for LLMs to securely query anonymized clinical data from a PostgreSQL-based Prontuário Eletrônico do Cidadão (PEC) database.

Setup Requirements

  • ⚠️Requires access to a specific 'Prontuário Eletrônico do Cidadão (PEC)' PostgreSQL database schema.
  • ⚠️Requires Python 3.10 or higher.
  • ⚠️Database connection details (host, port, name, user, password) must be configured via environment variables or a .env file.
Verified SafeView Analysis
The server demonstrates strong security practices: all database queries are read-only and use parameterized inputs to prevent SQL injection. It enforces strict limits on returned data rows (typically 50-200, up to 1000 for SOAP notes) and anonymizes sensitive patient identifiers (e.g., full names are converted to initials). Tools require specific filters to prevent broad data sweeps. Hardcoded database credentials are explicitly stated as defaults for development and users are instructed to override them with environment variables or a .env file for production. No 'eval' or other dynamic code execution patterns were found. The most sensitive data (SOAP notes) requires a specific patient ID and is subject to limits.
Updated: 2026-01-19GitHub→
0
0
Low Cost
pr0methevs icon

mcp-template

by pr0methevs

Sec9

This server provides a backend implementation for the Model Context Protocol (MCP) using Express.js, enabling real-time communication through Server-Sent Events (SSE) and managing tool execution.

Verified SafeView Analysis
The codebase appears well-structured and follows good practices for a template. It uses environment variables for configuration (PORT, HOST, SERVER_NAME, SERVER_VERSION), preventing hardcoded secrets. Input arguments for tools are validated against a schema, which helps prevent basic injection attacks related to tool arguments. Error handling for MCP messages is in place to avoid leaking internal server details. The default CORS configuration allows all origins, which is common for a template but should be restricted in a production environment. There are no obvious `eval` or dynamic code execution from user input for tool calls. The main minor concern is the lack of explicit connection limiting for SSE, which could be a denial-of-service vector if not managed at a proxy or application level in production.
Updated: 2025-12-09GitHub→
0
0
Low Cost
CodexVeritax icon

veritax_analytics

by CodexVeritax

Sec8

A cloud-agnostic analytics SDK for Model Context Protocol (MCP) servers, designed to track tool execution and server health events.

Setup Requirements

  • ⚠️Requires Python 3.10 or higher.
  • ⚠️A valid `api_key` is mandatory for client initialization.
  • ⚠️If loading configuration from a YAML file, `PyYAML` must be installed (`pip install pyyaml`).
Verified SafeView Analysis
The SDK utilizes Pydantic for robust input validation, which is a strong security practice. API keys are handled as standard bearer tokens. No direct 'eval' or 'exec' calls are found. The `file_path` validator automatically creates parent directories, which could be a path traversal vulnerability if the `file_path` is controlled by untrusted input, but for an SDK configured by a developer, this risk is mitigated. Metadata sanitization is basic (removes control characters) but appropriate for analytics data storage rather than direct rendering.
Updated: 2025-12-05GitHub→
0
0
Low Cost
Veritrust-VC icon

mcp-trust-framework

by Veritrust-VC

Sec2

The MCP Trust Framework provides an identity and trust layer for Model Context Protocol (MCP) servers, allowing AI agents and hosts to discover, identify, and verify the trustworthiness of external tools and data sources before interaction.

Setup Requirements

  • ⚠️Requires Python 3.10 or newer (for registry reference implementation).
  • ⚠️Registry uses in-memory storage; all data is lost on restart.
  • ⚠️Registry write operations (e.g., POST /mcp/servers) lack authentication and authorization, making it highly insecure for production.
  • ⚠️SDKs do not perform full cryptographic verification of Verifiable Credential signatures.
Review RequiredView Analysis
The provided reference implementation of the MCP Trust Registry is explicitly stated as 'for demonstration and testing' and 'does not implement authentication, cryptographic verification, or persistent storage.' **Critical Vulnerabilities in Reference Implementation (if used in production):** - **Authentication Bypass:** The `POST /mcp/servers` endpoint has no authentication, allowing any unauthenticated client to register, update, or potentially remove registry entries. This is a severe security risk, enabling malicious actors to inject untrusted servers or alter legitimate entries. - **Data Volatility:** It uses in-memory storage (`InMemoryStorage`), meaning all data is lost upon server restart. - **Weak CORS Policy:** Wildcard CORS (`allow_origins=["*"]`) is used, which is typically too permissive for production environments. **Limitations in SDKs:** - The Python and Node.js SDKs explicitly state they 'do not implement full cryptographic verification of Verifiable Credentials.' This is a major limitation for a 'trust framework,' as it means clients cannot cryptographically verify the integrity and authenticity of the credentials received, relying only on structural and simple policy checks. Full VC proof verification is left as a 'production exercise'. While the project includes a `security-considerations.md` outlining best practices, the reference implementation itself currently lacks these critical security controls, making it highly unsuitable for anything beyond isolated local development.
Updated: 2025-11-25GitHub→
PreviousPage 527 of 760Next