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
domdomegg icon
Sec9

Facilitates AI access to Google Maps Places API for searching locations, obtaining business information, and retrieving place photos for various applications like meeting coordination, local recommendations, and address validation.

Setup Requirements

  • ⚠️Requires manual creation and configuration of Google OAuth credentials (Google Cloud Project, enabled Places API (New), OAuth consent screen, and Web application client ID/secret).
  • ⚠️Requires specific environment variables: GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET for HTTP transport, or GOOGLE_ACCESS_TOKEN for stdio transport. Failure to provide them will result in server termination.
  • ⚠️The underlying Google Maps Places API is a paid service, incurring costs per request which are separate from any LLM token costs.
  • ⚠️The MCP_TRANSPORT environment variable must be explicitly set to 'http' for OAuth support; otherwise, it defaults to 'stdio' which requires a pre-obtained GOOGLE_ACCESS_TOKEN.
Verified SafeView Analysis
The server acts as an OAuth proxy, handling Google API credentials (Client ID/Secret) securely via environment variables. It does not store user access tokens, performing stateless token validation against Google's tokeninfo endpoint with caching to improve performance and gracefully handle expired tokens by returning HTTP 401, which is critical for client-side token refresh. Input/output schemas are rigorously validated using Zod, minimizing data-related vulnerabilities. No 'eval' or obfuscation is present, and network interactions are standard for an OAuth flow. The overall design prioritizes secure handling of external API access.
Updated: 2026-01-07GitHub→
0
0
Low Cost
EliseuSantos icon
Sec8

Provides a Micro-Copilot (MCP) HTTP server for AI agents to discover, query, and suggest React components from a centralized Storybook-documented design system in a monorepo architecture.

Setup Requirements

  • ⚠️Requires Node.js 18+ (recommended 20+)
  • ⚠️Requires pnpm 8+ (or npm/yarn with workspaces)
  • ⚠️Requires Docker for simulated production mode
  • ⚠️Requires Storybook static build to be run (`storybook:build`) before the MCP server can effectively provide component metadata.
  • ⚠️Requires VS Code or Cursor with MCP extension for AI interaction.
Verified SafeView Analysis
The server uses standard `express` practices. No `eval()` or code obfuscation was found. It relies on environment variables for sensitive paths/URLs, which is good practice. CORS is set to `*` (allow all origins), which is acceptable for a local/internal development tool but should be locked down to specific origins in a production deployment exposed externally. Input validation for tool arguments is basic string matching, which is sufficient for the current toolset, but should be carefully reviewed if new tools involving more complex operations (e.g., file paths, database queries) are added.
Updated: 2025-12-18GitHub→
0
0
Medium Cost
octbramantya icon

pfn-mcp

by octbramantya

Sec9

A conversational AI interface for industrial energy monitoring, analysis, and cost management using natural language queries against a PostgreSQL/TimescaleDB database.

Setup Requirements

  • ⚠️Requires an Anthropic API Key (Paid service for LLM inference).
  • ⚠️Requires Docker and Docker Compose for easy setup (though can be run natively).
  • ⚠️Requires a PostgreSQL database with TimescaleDB extension for time-series data.
  • ⚠️Requires Keycloak (or compatible OAuth2 provider) for authentication and user/tenant management.
Verified SafeView Analysis
The project demonstrates strong security practices with parameterized SQL queries preventing injection and robust Keycloak OAuth/JWT authentication for access control. Critical internal servers (e.g., SSE transport for the MCP protocol) are explicitly warned against public internet exposure in the source code. Custom formula parsing includes robust validation to prevent injection. Hardcoded secrets are avoided, relying on environment variables. The primary security consideration is ensuring proper deployment (e.g., internal network access only for certain services) and strong environment variable management.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
hariskarim41 icon

claude_prompter

by hariskarim41

Sec8

A universal Model Context Protocol (MCP) server for advanced prompt management, enabling structured AI workflows, hot-reloading, and integration with various AI clients.

Setup Requirements

  • ⚠️Requires Node.js 16+ to run.
  • ⚠️Full semantic analysis and LLM-driven chain features require an external LLM API key (e.g., OpenAI, Anthropic), which is a paid service.
  • ⚠️Requires an MCP-compatible client (e.g., Claude Desktop, Cursor Windsurf, Claude Code CLI) for interactive prompt management and execution.
Verified SafeView Analysis
The project demonstrates strong security awareness through explicit quality gates (e.g., `security-awareness.json` forbidding hardcoded secrets, input validation). The `CLAUDE.md` file strictly forbids direct file operations by the AI agent, mandating the use of secure MCP tools like `prompt_manager` and `SafeConfigWriter` for configuration changes. The use of Nunjucks for dynamic chain orchestration with LLM outputs is a high-risk area for template injection if not perfectly sanitized and escaped. While sanitization is mentioned, this remains an inherent complexity in such systems. Overall, the emphasis on security and controlled access is a significant positive.
Updated: 2025-12-03GitHub→
0
0
Low Cost
tspenov icon
Sec9

Provides an Elixir-based Model Context Protocol (MCP) server exposing BJJ position and video data as tools for AI assistants.

Setup Requirements

  • ⚠️Requires Elixir to be installed.
  • ⚠️This is a skeleton/library; real data providers must be implemented in a private application to replace the mock data.
  • ⚠️The server runs within an Elixir application (e.g., Phoenix or Plug) and requires specific router configuration.
Verified SafeView Analysis
The server itself is a skeleton using mock data, so direct security risks within this repository's code are minimal. No `eval` or obvious malicious patterns are present. The primary security considerations will arise from the private application's implementation of 'real' data providers (e.g., proper database access controls, input sanitization for queries, handling of sensitive video URLs or metadata). The current setup explicitly delegates data access to the consuming application, making this library's code itself quite safe.
Updated: 2025-11-27GitHub→
0
0
Low Cost
pentafive icon

your-spotify-mcp

by pentafive

Sec9

Connects AI assistants to Your Spotify for advanced music analytics and optional Spotify Web API control.

Setup Requirements

  • ⚠️Requires a self-hosted 'Your Spotify' instance to collect and store your listening history data.
  • ⚠️Tier 5 tools (Spotify Control) require separate Spotify Developer App credentials (Client ID, Client Secret, OAuth Access/Refresh Tokens) and will be disabled if not configured.
  • ⚠️Some Tier 4 Management tools (`analyze_affinity`, `rename_account`, `revoke_public_access`, `update_user_settings`) require an authenticated web session with Your Spotify, and cannot be used with a public token (read-only API limitation).
  • ⚠️The Your Spotify API typically limits data requests to 30 items, even if a higher limit is specified in a tool call.
Verified SafeView Analysis
The server uses environment variables for all sensitive API keys/tokens, which is a strong security practice. It communicates with AI clients via STDIO (standard input/output), eliminating local network listening vulnerabilities. HTTP requests to external APIs (Your Spotify and Spotify Web API) are handled securely with `axios` and rate limiting. Error messages are designed to be sanitized, preventing accidental exposure of sensitive information. The primary security considerations involve the user's secure management of environment variables and the underlying security of their self-hosted Your Spotify instance.
Updated: 2026-01-15GitHub→
0
0
Low Cost
BryanSchuetz icon

mcp

by BryanSchuetz

Sec9

This repository contains Markdown files detailing numerous international development projects, likely intended for an organizational website or public-facing project portfolio.

Verified SafeView Analysis
The provided 'SOURCE CODE' consists of static Markdown content files (.md), not executable server-side code. Therefore, traditional security risks typically found in server applications (e.g., 'eval' usage, obfuscation, direct network access, hardcoded API keys/secrets, or malicious code patterns) are not present within this content. The files do embed content via `iframe` tags from external sources (Flickr, Vimeo), which introduces a dependency on the security of those third-party platforms and the client-side rendering environment. However, the content itself is non-malicious.
Updated: 2025-12-08GitHub→
0
0
Low Cost
hgursel icon

NetContext-MCP

by hgursel

Sec6

AI-powered network automation and device management by enabling AI assistants to execute commands on network devices via Model Context Protocol (MCP).

Setup Requirements

  • ⚠️Requires access to live or simulated network devices (e.g., physical switches, GNS3, or Docker test servers).
  • ⚠️SSH host key verification is disabled by default, requiring manual configuration to enable for secure production use (vulnerable to MitM if not enabled).
  • ⚠️Manual configuration of client-specific JSON file (e.g., `~/.claude.json`) with absolute paths and credentials, which is prone to errors.
  • ⚠️GNS3 setup requires specific Aruba AOS-CX simulator image and significant RAM (8GB+).
Review RequiredView Analysis
The primary security risk is that SSH host key verification is disabled by default (`SSH_VERIFY_HOST_KEY=false` in environment config and `ssh.ts`), making the server vulnerable to Man-in-the-Middle (MitM) attacks. This is a critical flaw for a network management tool and must be manually enabled for production. While strong command sanitization (`rm`, `erase`, command chaining, path traversal, redirection) and file path validation for documentation are implemented, and YAML parsing uses `FAILSAFE_SCHEMA` to prevent code execution, the default SSH setting significantly lowers the overall security posture. Hardcoded test credentials exist in development configurations, emphasizing the need for secure environment variable management in production.
Updated: 2025-11-20GitHub→
0
0
Low Cost
bisVo159 icon
Sec3

A server for tracking and managing personal expenses via a FastMCP API.

Setup Requirements

  • ⚠️Requires 'uv' for dependency management, not 'pip'.
  • ⚠️Expense data is stored in the system's temporary directory, which may not be persistent across reboots or system cleanups, leading to potential data loss.
  • ⚠️Lacks any authentication or authorization, making it unsafe to expose on untrusted networks or use in multi-user environments.
Verified SafeView Analysis
The application critically lacks any authentication or authorization mechanisms. Any client capable of connecting to the server (which defaults to '0.0.0.0:8000') can perform all CRUD operations on expense data. This poses a severe privacy and security risk if the server is exposed to an untrusted network. Data is stored in the system's temporary directory, which can lead to data loss due to system cleanup processes. While SQL injection is mitigated by using parameterized queries, the complete absence of access control for sensitive financial data is a major vulnerability.
Updated: 2025-12-17GitHub→
0
0
Low Cost
Ce-terminals-02 icon
Sec1

Provides instructions or automation for deploying a Minecraft server on Amazon Web Services (AWS).

Setup Requirements

  • ⚠️Requires an AWS account and appropriate credentials for deployment.
Review RequiredView Analysis
A comprehensive security audit cannot be performed as no actual source code was provided beyond the README.md file. Without the implementation details, it's impossible to check for malicious patterns, 'eval' usage, obfuscation, network risks, or hardcoded secrets.
Updated: 2025-12-02GitHub→
0
0
Low Cost
web-dev-lab-official icon

mcp-server

by web-dev-lab-official

Sec7

Integrates AI applications with Google Calendar by exposing calendar functionalities as MCP tools, resources, and prompts.

Setup Requirements

  • ⚠️Requires Google Cloud Project with Calendar API enabled
  • ⚠️Requires `credentials.json` file from Google Cloud in the project root
  • ⚠️Requires Python 3.13+
  • ⚠️Requires `uv` package manager for setup and running
Verified SafeView Analysis
The server relies on `token.json` and `credentials.json` for Google OAuth, which contain sensitive `client_id`, `client_secret`, and refresh tokens. The `https://www.googleapis.com/auth/calendar` scope grants full access to the user's calendar. While standard for Google API integration, these files must be securely handled and protected to prevent unauthorized calendar access. The HTTP server (`server_http.py`) is exposed on localhost by default; exposure to public networks would require additional security measures. No malicious patterns or `eval` statements were found in the provided source code.
Updated: 2025-12-09GitHub→
0
0
Low Cost
ishant1234567890 icon

LLM_with_MCP_server

by ishant1234567890

Sec8

Facilitates real-time, AI-powered chat interaction between a Large Language Model and players within a Minecraft Pi Edition server.

Setup Requirements

  • ⚠️Requires a specific llama.cpp compatible LLM model file (e.g., ggml-model-q4_0.bin) placed at './models/7B/' (or specified via LLM_MODEL_PATH).
  • ⚠️Requires a running Minecraft Pi Edition server accessible to the script (default 'localhost:4711').
  • ⚠️Requires the 'llama_cpp' Python library, which can have complex installation requirements, especially for GPU support.
Verified SafeView Analysis
The server does not contain 'eval' calls, obfuscated code, or apparent hardcoded secrets. It connects to a local LLM and a Minecraft server, posting LLM responses to chat. The primary security consideration would be the content generated by the LLM if deployed in an uncontrolled environment, but the code itself does not introduce direct system vulnerabilities.
Updated: 2025-11-23GitHub→
PreviousPage 545 of 760Next