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(8554)

0
0
Medium Cost
RectangleEquals icon

godotmcp

by RectangleEquals

Sec7

AI-powered management of Godot game projects, enabling AI assistants to perform project management, file operations, git tracking, and debugging.

Setup Requirements

  • ⚠️Requires Windows 10/11 with WSL 2 and Docker Desktop with MCP Toolkit beta feature enabled.
  • ⚠️Requires Claude Desktop for Windows.
  • ⚠️Requires Godot 4.4+ with the GodotAI GDExtension (separate component) installed and configured.
Verified SafeView Analysis
The GodotMCP server acts as a proxy between Claude Desktop and the GodotAI GDExtension. It does not directly perform file or git operations, relying on GodotAI for these. The Python codebase itself appears clean, without 'eval', obfuscation, or hardcoded secrets. Network communication is local (127.0.0.1) for the MCP server and uses 'host.docker.internal' for connecting to GodotAI, suggesting an isolated Docker deployment. A critical security feature is the 'User Approval' workflow for all AI-suggested file modifications and deletions, which significantly mitigates risks. However, the overall system's security heavily depends on the GodotAI GDExtension (not provided here) and the user's diligence in reviewing changes. Potential risks include the powerful nature of an AI having control over a project, and the need for a secure GodotAI implementation.
Updated: 2025-11-28GitHub→
0
0
Medium Cost
daniel-iliesh icon

grocy-mcp

by daniel-iliesh

Sec4

Provides an LLM-friendly API layer over the Grocy REST API, enabling structured tools for managing stock, shopping lists, recipes, chores, and tasks within a Home Assistant environment.

Setup Requirements

  • ⚠️Requires Home Assistant instance for authentication and ingress routing.
  • ⚠️Requires Python 3.13 or newer.
  • ⚠️Dependencies include 'uvicorn' and 'websockets', requiring a local Python environment setup with these packages (or using `uv` runner).
  • ⚠️The `GROCY_API_URL` environment variable must include the Home Assistant ingress path (e.g., `http://homeassistant.local:9192/api/hassio_ingress/XXXX/api`).
Review RequiredView Analysis
The server uses a fully open CORS policy (`allow_origins=["*"]`), which is explicitly mentioned as being for 'browser-based MCP Inspector'. While this might be acceptable for a strictly local and isolated development/testing environment, exposing a server with such a permissive CORS policy to any network, especially the internet, presents a significant security risk. It could enable Cross-Site Request Forgery (CSRF) or data exfiltration from other websites. The authentication mechanism using Home Assistant's ingress session tokens and Grocy API keys from environment variables is good practice for credential management, but the open CORS undermines overall security for broader deployment. Errors from the Grocy API are also logged in detail, which could potentially expose sensitive information if not handled securely in production logs.
Updated: 2025-11-30GitHub→
0
0
Low Cost
kwangloc icon

mcp_server

by kwangloc

Sec8

This project demonstrates how to build and interact with a Micro-service Communication Protocol (MCP) server using the `fastmcp` library, exposing callable tools over HTTP.

Setup Requirements

  • ⚠️Requires `fastmcp` library
  • ⚠️Requires `httpx` library
  • ⚠️Client examples default to a remote FastMCP instance; local testing requires client URL modification (e.g., to 'http://localhost:8080/mcp').
Verified SafeView Analysis
The server listens on `0.0.0.0` by default, which is common for cloud deployments but requires external network security measures. No obvious hardcoded secrets or dangerous functions like `eval` are present. External API calls are made to `api.frankfurter.app`, which is a legitimate currency exchange service.
Updated: 2026-01-19GitHub→
0
0
Low Cost
jackuh105 icon

macau-weather-mcp

by jackuh105

Sec9

Provides real-time and forecast weather data for Macau from the Macau Meteorological and Geophysics Bureau (SMG) to AI assistants via Model Context Protocol (MCP) tools.

Setup Requirements

  • ⚠️Requires Python 3.12 or higher.
  • ⚠️Dependencies (`httpx`, `mcp[cli]`) must be installed using `uv` (recommended) or `pip`.
Verified SafeView Analysis
The server fetches data from trusted government XML APIs (smg.gov.mo) and uses standard libraries for HTTP requests (httpx) and XML parsing (ElementTree). There is no use of `eval` or obfuscation. The `docker-compose.yml` file includes an optional MCPO bridge service with a placeholder `api-key "top-secret"`, which users should replace with a strong secret if deploying MCPO for production use. The core MCP server itself does not handle sensitive secrets.
Updated: 2025-11-28GitHub→
0
0
Low Cost
dancing-clown icon
Sec9

Provides an Ethereum Model Context Protocol (MCP) server for AI agents to safely query blockchain data like ETH/ERC20 balances, token prices, and simulate Uniswap token swaps without broadcasting actual transactions.

Setup Requirements

  • ⚠️Requires Rust 1.91.1+ toolchain (rustup.rs)
  • ⚠️Requires an Ethereum RPC endpoint URL (e.g., Infura, Alchemy, or public ones like llamarpc.com)
  • ⚠️For advanced features like actual transactions (not implemented in provided code), a private key would be needed securely.
Verified SafeView Analysis
The server explicitly prioritizes safety by using `eth_call` for all `swap_tokens` simulations, ensuring no actual transactions are broadcasted to the blockchain. Read operations (balance, price) do not require a private key. When a `PRIVATE_KEY` is provided, it is loaded from environment variables, which is good practice. Public contract addresses are hardcoded, which is expected. Overall, the design focuses heavily on preventing real asset loss during simulations and read-only operations.
Updated: 2025-11-29GitHub→
0
0
Medium Cost
aosyang icon

TaskMCP

by aosyang

Sec4

A multi-workspace task management system featuring hierarchical tasks, drag-and-drop sorting, real-time synchronization, and AI agent integration via an MCP server for natural language operations.

Setup Requirements

  • ⚠️Requires a local Ollama server running and the `qwen3:14b` model (or configured alternative in `agent_config.toml`) downloaded.
  • ⚠️Python versions older than 3.11 require the `tomli` library for TOML configuration files.
  • ⚠️The Flask web server uses a hardcoded `SECRET_KEY`, which is a security risk for any non-development environment and needs to be addressed.
Verified SafeView Analysis
The Flask web server uses a hardcoded `SECRET_KEY` ('task-secret-key'), which is a critical security vulnerability, especially if the application were to be deployed in a production environment or exposed to the internet, as it compromises session security. The `server_config.toml` allows binding to `0.0.0.0`, which could expose the application over the network without explicit authentication or authorization for web and API endpoints, posing a risk to task data. Frontend uses DOMPurify for Markdown sanitization, which is good for preventing XSS. Workspace names are validated. The Telegram bot enforces an optional user ID whitelist, which is a good practice.
Updated: 2026-01-19GitHub→
0
0
Low Cost
Blurjp icon
Sec9

Processes Figma design links into optimized, tiled images suitable for vision models like Claude.

Setup Requirements

  • ⚠️Requires a Figma Personal Access Token with file read access.
  • ⚠️Node.js version 18 or higher is required.
  • ⚠️The server must be running continuously to process requests (registration with Claude is separate from starting the server).
Verified SafeView Analysis
The server implements robust security measures including in-memory storage of Figma tokens per session (cleaned after 1 hour), strict file permissions (0o600 for token file, 0o700 for directory), and a rate limiter (100 req/min per IP). The use of 'undici' for network requests and 'sharp' for image processing are standard and generally secure libraries. File system operations for output ('./out/figma/') are expected for its functionality; while the 'out_dir' parameter could theoretically be a vector for directory traversal if maliciously crafted, standard path resolution helps mitigate typical risks. No 'eval' or obvious obfuscation detected. The multi-tenant token handling stores tokens in memory only, enhancing security for public deployments.
Updated: 2026-01-18GitHub→
0
0
Medium Cost
roizenlabs icon

sportintel-mcp

by roizenlabs

Sec9

AI-powered sports analytics for Daily Fantasy Sports (DFS), betting, and AI agents, providing explainable player projections, lineup optimization, and live odds aggregation.

Setup Requirements

  • ⚠️Requires BallDontLie API key (GOAT tier, $39.99/month minimum for '/stats' endpoint).
  • ⚠️Requires The Odds API key for 'get_live_odds' tool (free tier available but limited).
  • ⚠️Requires HuggingFace API key for AI injury risk detection (free tier available).
Verified SafeView Analysis
API keys are managed via environment variables, which is a good security practice. The Playwright-based browser scraper for DraftKings is commented out for MVP; if enabled, proper sandboxing would be crucial to mitigate potential remote code execution risks inherent with browser automation. No direct 'eval' or arbitrary shell command execution from user input is apparent in the server's core logic.
Updated: 2025-12-01GitHub→
0
0
High Cost
corbyjamesibm icon

spm-mcp

by corbyjamesibm

Sec9

Manage Targetprocess projects, track work, and update tasks through natural language conversations with an AI assistant.

Setup Requirements

  • ⚠️Requires Docker installed and running for containerized deployment.
  • ⚠️Requires a Targetprocess account with domain, credentials (username/password or API key), and API access permissions.
  • ⚠️Specific Targetprocess API permissions are needed for various operations (e.g., creating/updating entities).
  • ⚠️The `analyze_attachment` tool can generate very large base64 encoded outputs (up to 50MB file size, potentially millions of tokens) which will incur significant LLM token costs if this data is sent to the AI model.
Verified SafeView Analysis
The server handles Targetprocess credentials via environment variables or config files, which is a standard secure practice. Input validation for tool parameters is implemented using Zod. The `QueryBuilder` includes custom parsing and escaping for SQL injection prevention in the Targetprocess query language. The `AnalyzeAttachmentOperation` explicitly implements strong security measures including file size limits (50MB), MIME type whitelisting, and suspicious filename pattern detection before downloading and base64-encoding content. While base64 encoding large files (up to 50MB) means potentially massive token consumption by the LLM, the `AnalyzeAttachmentOperation` does so under strict security controls, making the process itself safe from the server's perspective.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
SynTaek icon
Sec9

Provides LLMs with persistent memory of coding history by tracking code evolution in a shadow git repository, enabling session continuity and flow-based development.

Setup Requirements

  • ⚠️Requires `git` command-line tool to be installed and in the system PATH.
  • ⚠️Requires `uv` package manager for installation and `uvx` for running directly from the Git URL as configured for AI clients.
  • ⚠️The target project directory must be a Git repository (initialized with `git init`).
  • ⚠️Requires Python 3.14 or newer.
Verified SafeView Analysis
The server primarily interacts with the local filesystem and a hidden Git repository (.trajectory) via the gitpython library. It does not contain direct use of 'eval', obfuscation, or hardcoded secrets. Network interactions are confined to the local MCP server communication. The primary security consideration is the potential for a compromised AI client to instruct the server to track arbitrary paths, though the system's design (requiring a local `git init` and tracking only specified project paths) limits broader system-level exploitation from the MCP server's side itself.
Updated: 2025-12-05GitHub→
0
0
Low Cost
vikkysarswat icon

trade_detail

by vikkysarswat

Sec8

Display real-time stock prices and market index data as interactive widgets within ChatGPT.

Setup Requirements

  • ⚠️Requires Node.js 18+ and Python 3.10+ to set up and run locally.
  • ⚠️The `BASE_URL` environment variable is critical and must be explicitly set to the deployed application's URL for widgets to render correctly (e.g., `https://your-app-name.onrender.com`).
  • ⚠️For local development, two separate processes are required: one for the frontend assets (`pnpm run serve`) and one for the Python backend (`uvicorn server.main:app`). In some deployment environments (like Render), assets might be served implicitly, but explicit static file serving might be needed depending on the setup (e.g., Nginx for self-hosting).
Verified SafeView Analysis
The server uses `CORSMiddleware` with `allow_origins=["*"]`, which permits requests from any origin. While common for public APIs, for enhanced security, it is generally recommended to restrict this to known domains if possible. No direct 'eval' or hardcoded secrets were identified. The `BASE_URL` environment variable is critical for correct widget rendering in deployment and must be configured securely.
Updated: 2025-11-28GitHub→
0
0
Low Cost
Snehalaxmi-S icon

mcp-server

by Snehalaxmi-S

Sec9

This server acts as an API Gateway, aggregating multiple external OpenAPI-defined services (like Petstore and Tenable) into a single unified endpoint using the FastMCP framework.

Setup Requirements

  • ⚠️Requires the 'fastmcp' Python library to be installed (e.g., `pip install fastmcp`).
  • ⚠️Relies on external OpenAPI specification URLs (Petstore, Tenable) being accessible for server setup.
  • ⚠️Default port 8000 must be available for the server to bind to.
Verified SafeView Analysis
The code itself is clean, uses standard libraries (httpx, asyncio, fastmcp), and does not contain obvious malicious patterns, hardcoded secrets, or dangerous functions like 'eval'. It binds to '0.0.0.0' for the main server, which is common but means it listens on all network interfaces. The primary security consideration would be the integrity and trustworthiness of the external OpenAPI specifications it fetches and aggregates, and the security of the backend APIs it proxies to.
Updated: 2025-11-24GitHub→
PreviousPage 542 of 713Next