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

m95-mcp

by dbshadow

Sec4

Manages M95 network devices, controlling functionalities and exposing them as tools for automated agents and AI assistants via a FastMCP server.

Setup Requirements

⚠️Requires a `.env` file with M95_HOST, M95_USER, and M95_PASSWORD configured.
⚠️Disables SSL certificate verification (`verify=False`) for device communication, posing a security risk (MITM vulnerability).
⚠️Requires Python 3 and dependencies listed in `requirements.txt`, including the `fastmcp` library.

Review RequiredView Analysis

The `m95_api.py` explicitly disables SSL certificate verification (`verify=False`) for HTTPS requests to the M95 device. While this might be intended for devices with self-signed certificates, it severely compromises transport security, making the communication vulnerable to Man-in-the-Middle (MITM) attacks. This could allow attackers to intercept or alter commands and data. Credentials (M95_PASSWORD) are loaded from a `.env` file, which is good practice, but their protection relies on the file system permissions. The MCP server listens on `0.0.0.0:8444`, meaning it's accessible from all network interfaces; ensure proper network segmentation if this server is exposed beyond a trusted local network.

Updated: 2025-12-02GitHub→

Low Cost

credential-free

by skutry

Sec9

A local MCP server and CLI tool for static analysis to detect hardcoded API keys, tokens, and credentials in source code and files.

Setup Requirements

⚠️Requires Python 3.10+
⚠️Optional dependencies (tree-sitter, tree-sitter-python, tree-sitter-javascript) are needed for 'deep' scanning profile's AST analysis.

Verified SafeView Analysis

The codebase demonstrates good security practices for a tool processing arbitrary user files. It actively prevents path traversal, limits file and archive member sizes to mitigate resource exhaustion and regex DoS attacks on long lines, and does not use dangerous functions like 'eval'. No hardcoded secrets for its own operation were found. The tool's scanning logic is local, minimizing external attack surface.

Updated: 2025-12-01GitHub→

Low Cost

incident-timeline-mcp

by ian-de-marcellus

Sec9

Extracts structured information from incident response logs to aid in timeline analysis, action tracking, entity identification, and severity assessment.

Setup Requirements

⚠️Requires Claude Desktop to be installed and running to use as an MCP server.
⚠️The Claude Desktop configuration requires absolute file paths for the Python interpreter and the server.py script.
⚠️Requires a Python 3 virtual environment and dependency installation (`pip install -r requirements.txt`).

Verified SafeView Analysis

The server primarily relies on Python's regex engine and string manipulation for text extraction. It runs as a local MCP server, communicating via stdio, which inherently limits network exposure. There are no direct calls to external APIs, use of 'eval' with untrusted input, or hardcoded secrets observed in the provided source code. Input sanitization is handled via context-based filtering for regex patterns (e.g., timestamps, actors), which helps mitigate false positives. The main potential (minor) risk would be if complex, malicious regex input could trigger a ReDoS vulnerability, but this is less impactful in a local, single-user context.

Updated: 2025-11-24GitHub→

Medium Cost

mcppractice

by Rayaannotfound

Sec8

This MCP server provides tools for an AI agent to fetch cryptocurrency news and prices, send emails, and create tasks in an external system.

Setup Requirements

⚠️Requires an SMTP server configured with credentials for sending emails.
⚠️Requires an external 'TASK_API_BASE' endpoint for task creation and a corresponding 'TASK_API_TOKEN'.
⚠️Requires Python packages 'fastmcp', 'requests', and 'feedparser' to be installed.

Verified SafeView Analysis

The code uses environment variables for all sensitive information (SMTP credentials, API tokens, API base URLs), which is a good security practice. It uses standard, secure libraries for network operations (requests with timeout, smtplib with SSL/TLS context). No 'eval' or other dynamic code execution is present. External API calls (CoinGecko, Google News RSS, custom task API) are made over HTTPS. The 'Trust Vox Tech' branding within the email footer and advice section, while part of the agent's personality, could be considered a minor social engineering aspect to be aware of in a real-world deployment.

Updated: 2025-11-20GitHub→

Medium Cost

examples

by Cronos402

Sec7

A Next.js application demonstrating how to build an OpenAI Apps SDK compatible MCP server with widget rendering in ChatGPT.

Setup Requirements

⚠️OpenAI developer mode is required to connect the MCP server to ChatGPT.
⚠️The `auth-example` requires an external authentication service URL (`NEXT_PUBLIC_AUTH_URL`).
⚠️The `cronos-weather-server` requires configuring a Cronos wallet address for payment reception and the Cronos402 CLI for testing paid tools.

Verified SafeView Analysis

The application uses broad CORS headers (`Access-Control-Allow-Origin: *`) and patches browser APIs (`history.pushState`, `window.fetch`, `MutationObserver` for `<html>`) to function within a ChatGPT iframe. While these are necessary for the specific integration, they inherently increase the attack surface and complexity compared to a standard web application. The `baseURL` is derived from Vercel environment variables, reducing direct user manipulation risk. The `auth-example` relies on an external authentication service via `NEXT_PUBLIC_AUTH_URL` and forwards cookies for session management. No obvious malicious code patterns or critical hardcoded secrets were identified in the provided snippets for the server operations.

Updated: 2026-01-18GitHub→

Low Cost

mcp-gitlab-mcp-server2

by bioanywhere

Sec8

This repository provides Docker deployment configuration for an MCP (Managed Cloud Platform) server, enabling its build and execution.

Setup Requirements

⚠️Docker Desktop or Engine required for local development and deployment.
⚠️Authentication to GitHub Container Registry may be required if the image is private.

Verified SafeView Analysis

The provided code consists of Docker deployment configurations and shell scripts. No direct application logic is available for review. The scripts themselves use standard Docker commands and do not contain 'eval', obfuscation, or hardcoded secrets. Network exposure is limited to port 3000, which is typical for a web server. The security score reflects the safety of the deployment mechanism, not the unseen application code within the Docker image.

Updated: 2025-12-01GitHub→

Medium Cost

ags-api-mcp-server

by AccelByte

Sec9

Provides AI assistants with access to AccelByte Gaming Services APIs through OpenAPI integration.

Setup Requirements

⚠️Requires AccelByte OAuth Client configuration in Admin Portal, with a Redirect URI that exactly matches the server's configured URI (e.g., http://localhost:3000/oauth/callback).
⚠️Docker is the recommended deployment method for easy integration with AI Assistants, implying it's a necessary component for the intended quick start setup.
⚠️The 'openapi-specs' directory must be populated with valid OpenAPI definitions, which may require a separate process or source for users beyond the initial setup.

Verified SafeView Analysis

The server employs several good security practices: environment variables for configuration, token masking in logs, PKCE for OAuth flows, OIDC token verification with JWKS, Helmet for HTTP security headers, and rate limiting (in V2). It adheres to a stateless architecture in V2, reducing session-related attack surfaces. A minor concern in V1's `StreamableHTTPTransport` is its `validateOrigin` which logs warnings but still allows non-localhost origins by default, which should be tightened for production outside of specific local scenarios. Furthermore, while the server uses JSONPath expressions, care must be taken when exposing these to LLMs to prevent potential injection if user input directly influences the expression, although the current implementation doesn't appear to directly facilitate this risk.

Updated: 2026-01-14GitHub→

Medium Cost

facebook-ads-mcp

by patpil-cloudflare-mcp

Sec9

Production-ready Cloudflare MCP server for Facebook Ad Library competitive intelligence, combining Apify scraping with AI-powered strategic analysis and format aggregation.

Setup Requirements

⚠️Requires configuration of multiple Cloudflare bindings (D1, KV, Durable Objects) and external API keys (WorkOS, Apify, Cloudflare AI Gateway).
⚠️An active Apify account is mandatory, along with the `apify/facebook-ads-scraper` Actor configured for use.
⚠️A manual code change is required for Zod v4 compatibility (`z.string().url()` to `z.url()`) in `src/server.ts` and `src/api-key-handler.ts` as detailed in `MIGRATION_PLAN.md`.

Verified SafeView Analysis

The server implements robust security practices: OAuth 2.1 with PKCE using WorkOS for user authentication, API key authentication with SHA-256 hashing and user-level validity checks against a D1 database, and explicit PII redaction and output sanitization via `pilpat-mcp-security` (v1.1.0+) in all tool outputs (Step 4.5). Secrets are managed through `.dev.vars` locally and `wrangler secrets` in production, not hardcoded. Concurrency is controlled via a Durable Object semaphore to prevent resource exhaustion. URL validation is present before external API calls. No `eval` or suspicious dynamic code execution is observed. The choice to disable email/phone PII redaction for ad content is a documented, deliberate business decision for publicly available marketing data.

Updated: 2025-12-28GitHub→

Low Cost

kubeagentics-catalog

by kubeagentics

Sec6

Provides a curated, machine-readable catalog of Kubernetes-relevant Model Context Protocol (MCP) servers, facilitating their discovery and one-click installation within the KubeAgentics platform.

Setup Requirements

⚠️Requires Node.js and npm/npx to execute the described MCP servers.
⚠️Most MCP servers require specific environment variables (e.g., API keys, URLs) that must be provided by the user.
⚠️Relies on external npm packages which must be accessible from the network where KubeAgentics executes them.

Verified SafeView Analysis

The catalog itself is a static JSON file and inherently safe to process. However, the MCP server definitions within it specify 'command' fields that use 'npx' to execute external npm packages (e.g., '@anthropic/mcp-server-prometheus'). This means the consuming application (KubeAgentics) would download and run arbitrary code from npm. This introduces a supply chain security risk if a listed npm package is compromised or if a malicious entry is added to the catalog, as the executed package would have full access to the host environment. No 'eval' or obfuscation is present in this repository's own files, nor are any secrets hardcoded for the catalog's operation.

Updated: 2025-12-01GitHub→

High Cost

MCP-Project

by MonkeyDLuffy-07

Sec5

Provides a scalable and secure API gateway for multiple Large Language Models, targeting the Indian developer market with features like API key authentication and tiered pricing.

Setup Requirements

⚠️Requires API keys for upstream LLM providers (e.g., OpenAI, Anthropic) to function.
⚠️Requires a database to manage users, API keys, and track usage for tiered pricing.
⚠️Python environment and FastAPI dependencies need to be installed for local development/deployment.

Review RequiredView Analysis

Cannot perform a full security audit without access to the codebase. Based solely on the README, the mention of 'API key authentication' indicates an intent for security. Specific checks for 'eval', obfuscation, or detailed network risks are not possible with the provided information.

Updated: 2025-11-17GitHub→

Low Cost

xiaozhi-mcp-server

by syncaster

Sec2

Extending AI model capabilities by exposing tools for remote control, calculations, and external system interactions.

Setup Requirements

⚠️Requires `MCP_ENDPOINT` environment variable to connect to the WebSocket endpoint.
⚠️Requires `mcp_config.json` for managing multiple tool servers, or a direct script path.
⚠️The hardcoded JWT token in `xiaozhi_chat_web.py` would need to be replaced with a secure method (e.g., environment variable).

Review RequiredView Analysis

The primary security risk is the use of `eval()` in the `calculator` tool implementations (`calculator.py` and `server.py`). These tools directly execute arbitrary Python expressions provided by the AI model. While `calculator.py` attempts to sandbox `eval` to `math` and `random` modules, this is often insufficient to prevent sophisticated arbitrary code execution. The `server.py` example's `eval(python_expression)` is even more dangerous, offering no sandboxing at all. This allows an attacker (or a malfunctioning AI) to execute arbitrary code on the server where the tool is running, potentially leading to system compromise. Additionally, `xiaozhi_chat_web.py` contains a hardcoded JWT token in `XIAOZHI_ENDPOINT`, which is a critical secret management vulnerability.

Updated: 2026-01-18GitHub→

Low Cost

ethereum-mcp-server

by cowboy0015

Sec9

Provides an MCP (Multi-Modal Agent Protocol) server for interacting with the Ethereum blockchain, offering tools for querying balances, fetching token prices, and building Uniswap V3 swap transactions.

Setup Requirements

⚠️Requires a valid Ethereum private key (with sufficient ETH for gas) to be set as the `PRIVATE_KEY` environment variable.
⚠️Requires a Rust toolchain to build and run the application.
⚠️Interaction with the Ethereum blockchain means real transaction costs (gas fees) for swap operations (though the current implementation primarily builds and simulates transactions).

Verified SafeView Analysis

The server loads sensitive `PRIVATE_KEY` from environment variables, which is good practice. A hardcoded private key exists in `Config::ethereum_default()` but is explicitly for testing and not used in the main application flow. Users must securely manage their `PRIVATE_KEY` environment variable, as it's used for transaction signing. No 'eval', obfuscation, or other critical malicious patterns were found.

Updated: 2025-12-03GitHub→

PreviousPage 367 of 760Next