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

MemoryGraph

by DanNsk

Sec8

Visualizes knowledge graphs from the multi-memory-mcp project in an interactive 3D force-directed graph.

Setup Requirements

⚠️.NET 8.0 SDK is required.
⚠️SQLite databases must adhere to a specific schema and be located in the configured 'MemoryFolderPath' (default: ./.aim/). Although sample databases are generated in Development mode, custom data requires manual setup.

Verified SafeView Analysis

The application handles graph data loading via API calls where the 'database' parameter is URL-encoded on the frontend. A critical security aspect in the backend (MemoryGraphService/SqliteDataService, not fully provided) is rigorous path validation to prevent directory traversal when accessing SQLite files, ensuring only files within the configured 'MemoryFolderPath' are accessed. Frontend HTML escaping is used to mitigate XSS risks. No 'eval' or obvious hardcoded secrets were found in the provided code snippets.

Updated: 2025-12-03GitHub→

Medium Cost

bybit-ai-trader

by TheMacroeconomicDao

Sec3

AI-driven cryptocurrency market analysis, signal generation, and automated trading operations on Bybit, with Telegram integration for signal delivery and a WebUI for monitoring.

Setup Requirements

⚠️Requires valid Bybit API Key and Secret with appropriate read/trade permissions.
⚠️Requires OpenRouter.ai API Key (`QWEN_API_KEY`) and sufficient credits for Qwen models.
⚠️Requires Telegram Bot Token (`TELEGRAM_BOT_TOKEN`) and specific chat IDs (`TELEGRAM_CHAT_IDS`) for notifications.
⚠️Python dependencies must be installed via `pip install -r requirements.txt`.
⚠️WebUI setup and server (`bybit-analysis`) require Node.js and pnpm.

Review RequiredView Analysis

The `CURSOR_MCP_CONFIG.json` provided directly exposes `BYBIT_API_KEY` and `BYBIT_API_SECRET` within its `env` block. While the system's `load_credentials` function prioritizes environment variables, this configuration itself is a critical vulnerability if exposed. The project's historical records (e.g., `archive/fixes/FIX_PROMPTS.md`) also detail past leaks of multiple API keys and a Telegram token, with explicit instructions for Git history cleansing, indicating a significant past security lapse. The `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_IDS` are also explicitly mentioned in plain text in documentation files, which is sensitive information.

Updated: 2025-12-19GitHub→

High Cost

mcp-server

by leadfuze

Sec9

Enable AI agents to enrich contacts and companies with verified business data like emails, phone numbers, and job titles for sales prospecting.

Setup Requirements

⚠️Requires a LeadFuze API Key (LeadFuze is a paid service, credits consumed per match/validation)
⚠️Requires Node.js 18.0.0 or higher to run locally
⚠️Self-hosting requires `npm install` and `npm run build`

Verified SafeView Analysis

The server handles API keys via environment variables or Authorization headers, which is standard and secure for its purpose. CORS is properly configured to restrict access to allowed origins (e.g., Claude.ai, localhost). No `eval` or direct file system access beyond module loading. Sessions are managed to ensure per-user API key isolation. Error handling in the client provides user-friendly messages, though the raw API error text could be verbose if the upstream LeadFuze API were to return sensitive data in an error payload (this is not a direct vulnerability of this server's code, but a general consideration for API proxies).

Updated: 2025-12-14GitHub→

Low Cost

ez-mcp-server

by dbrooks61785

Sec8

Provides a simple server implementation of the Model Context Protocol (MCP) using Bun and Express, featuring a basic ping tool for automation contexts.

Setup Requirements

⚠️Requires Bun runtime (v1.3.3 or compatible) to execute the Node.js server.
⚠️The `README.md` specifies `bun run index.js` which only prints 'Hello via Bun!' and does not start the actual MCP server. The correct command to start the MCP server is `bun run server.js`.
⚠️The `manifest.json` describes a `/ping` tool with an `endpoint` of `http://localhost:3030/ping`. While `server.js` starts an MCP server on `localhost:3030` and internally defines a 'ping' tool, the primary interaction is via `/mcp`. A separate `mcp_server.py` file *does* expose a `/ping` endpoint, suggesting it might be an expected external service or a potential misconfiguration/redundancy.

Verified SafeView Analysis

The project uses `dotenv` for environment variable loading, which is a good practice for managing configurations and secrets, though no specific environment variables are explicitly referenced in the provided `server.js` code. The core server functionality relies on the `@modelcontextprotocol/sdk`, whose security would depend on its own implementation. An Express server is used, which is a mature framework. No `eval` or obvious obfuscation is present. The presence of a separate `mcp_server.py` (FastAPI) in the repository is ambiguous but is likely a separate component or example not directly integrated with the Bun server.

Updated: 2025-12-03GitHub→

Medium Cost

weather_mcp_adk_python

by Lovee93

Sec7

An AI agent that provides weather forecasts for US locations by geocoding user input, fetching weather data, and optionally sending activity recommendations to an AWS SNS topic.

Setup Requirements

⚠️Requires `GOOGLE_MAPS_API_KEY` environment variable set (Google Maps API is typically a paid service).
⚠️Requires AWS credentials configured for the `personal` profile with permissions to publish to `arn:aws:sns:ap-southeast-2:123456789:weather` (or a similar SNS topic).
⚠️The Node.js-based `weather_mcp_server` must be built (`npm install && npm run build` in `weather_mcp_server` directory) and its `build/index.js` path correctly specified in `weather_adk_python/agent.py`.
⚠️Requires Node.js, npm/uvx/npx to be installed on the host system to run the MCP servers.

Verified SafeView Analysis

The agent relies on `npx` and `uvx` to download and execute external MCP servers (`@modelcontextprotocol/server-google-maps` and `awslabs.amazon-sns-sqs-mcp-server@latest`) at runtime. While these appear to be official sources, dynamically executing external packages always carries a supply chain risk if the packages themselves were to be compromised. Input validation for latitude/longitude is performed using Zod in the local weather server. Environment variables are used for sensitive keys, which is good practice. The `AWS_PROFILE` is hardcoded to 'personal', relying on pre-configured AWS credentials on the host machine.

Updated: 2026-01-19GitHub→

Low Cost

claude-code-safety-net

by waleedkhanbaloch

Sec9

The application acts as a safety shield for AI coding agents, preventing the execution of harmful Git and filesystem commands to protect user data and systems.

Setup Requirements

⚠️Requires Bun runtime and package manager for development and local execution of scripts.
⚠️Requires Node.js and npm/npx for `cc-safety-net` CLI wrapper execution (e.g., within Claude Code hooks).
⚠️Designed to run as a plugin/hook within AI coding environments like Claude Code or OpenCode.

Verified SafeView Analysis

The project is explicitly designed for security, implementing defense-in-depth measures such as strict and paranoid modes, recursive command analysis for nested shell calls (e.g., `bash -c`), detection of dangerous interpreter one-liners, and audit logging with secret redaction. Input parsing for hooks uses `JSON.parse` and tokenization relies on `shell-quote`, which is used with precautions (e.g., `ENV_PROXY` to prevent premature variable expansion). Path sanitization is performed for audit log filenames. While shell parsing is inherently complex and can hide subtle bypasses, the comprehensive set of rules and defensive coding practices indicate a strong focus on safety. The plugin operates locally and does not introduce external network risks during command analysis.

Updated: 2026-01-19GitHub→

Medium Cost

chi311-mcp-server

by changju784

Sec3

Automate the submission of Chicago 311 service requests via web browser automation, exposed as an API for LLM agents.

Setup Requirements

⚠️Requires Python 3.12+ and Playwright browser installation (`python -m playwright install`).
⚠️By default, form submissions run in 'dry-run' mode; enabling full submission requires setting specific environment variables (`CHI311_FULL_SUBMIT=true`), which is explicitly warned as dangerous without further security measures.
⚠️Default CORS policy (`allow_origins=["*"]`) makes the server unsafe for public exposure without modification and adding API key protection.

Review RequiredView Analysis

The server uses `allow_origins=["*"]` for CORS, which is a significant security risk for public deployment. Critical endpoints like `/mcp/submit_311_request`, `/mcp/tools/fetch`, and `/sse/broadcast` are unauthenticated by default. The `/mcp/tools/fetch` endpoint can expose module source code and JSON schemas publicly. The `submit_311_request` endpoint, if `CHI311_FULL_SUBMIT` is enabled, can trigger actual form submissions without default authentication or rate-limiting, posing a risk of abuse. The README explicitly warns about these risks but the default configuration does not mitigate them.

Updated: 2025-12-12GitHub→

Medium Cost

mcp-youtube

by Steve-651

Sec8

Extracts YouTube video transcripts and metadata for AI analysis within Claude Desktop.

Setup Requirements

⚠️Requires yt-dlp to be installed and available in your system PATH.
⚠️Requires Node.js v16.0.0 or later.
⚠️Transcripts are saved to a local folder (defaults to `~/youtube-transcripts`), which is configurable by the user.

Verified SafeView Analysis

The server uses `child_process.execFile` to invoke `yt-dlp`. This is safer than `exec` as it prevents shell injection by design, and arguments are managed internally. No hardcoded secrets or direct 'eval' calls were found. The primary security consideration is the reliance on the externally installed `yt-dlp` utility and the user-configurable `TRANSCRIPTS_FOLDER` which could be set to a sensitive location by the user, though the application itself does not expose these files over a network.

Updated: 2025-12-15GitHub→

Medium Cost

AI-Challenge-mcp-server

by sobolvb

Sec5

This project is a Minecraft Coder Pack (MCP) server designed to host an AI challenge, enabling automated agents to interact within a customized Minecraft environment.

Setup Requirements

⚠️Requires Java Development Kit (JDK) 8+ (likely 17+ for modern Minecraft/Gradle projects).
⚠️Requires network access during initial setup to download Gradle dependencies.
⚠️Requires a compatible Minecraft client or AI agent to interact with the server's challenge environment.

Review RequiredView Analysis

Cannot conduct a full security audit without access to the source code in the `src` directory. As a server application, it will open network ports; security depends entirely on the implementation of input validation, authentication, and authorization within the application logic. No obvious 'eval' or code obfuscation found from the file list. Potential network risks exist if interfaces for AI agents are not robustly secured.

Updated: 2025-11-17GitHub→

Low Cost

mcp-server-docker

by hanydd

Sec7

A Dockerized platform for orchestrating and exposing multiple Python-based Micro-Agent servers (FastMCP tools) through a single Nginx reverse proxy.

Setup Requirements

⚠️Docker required
⚠️Relies on the mcp.server.fastmcp framework
⚠️Dynamically modifies Python scripts and Nginx configuration at startup

Verified SafeView Analysis

The system dynamically modifies Python scripts via AST manipulation and executes them in separate processes. It also generates and runs Nginx configuration based on scanned filenames. While performed on internal, presumably trusted, source files, this dynamic execution environment requires careful management of the `/app/mcp` directory contents. Open CORS headers (`Access-Control-Allow-Origin: *`) are used. No 'eval' or hardcoded secrets found.

Updated: 2025-11-19GitHub→

Low Cost

mcp-app-python

by ecociel

Sec8

Develop and test custom OpenAI (ChatGPT) App Widgets, demonstrating both plain HTML and React-based UI components served by a minimal Python MCP server.

Setup Requirements

⚠️Requires Python 3.x and specific packages (`mcp`, `fastmcp`, `uvicorn`, `pydantic`) installed via pip.
⚠️Requires Node.js and npm for building the React widget. The `npm run build-widgets` command must be executed before running `server.py`.
⚠️For testing with ChatGPT, the local server needs to be exposed to the public internet using a tool like ngrok.

Verified SafeView Analysis

The Python server code (both server.py and server-html.py) appears robust, handling file reads and HTTP requests without obvious direct vulnerabilities like `eval` or command injection. The `load_html` function in `server.py` involves string manipulation to inline JS/CSS and move script tags for self-contained widget embedding; while this carries an inherent, albeit small, risk if not perfectly implemented, it's for known build artifacts. Client-side widgets are expected to read `window.openai.toolOutput`; developers should ensure proper sanitization/escaping if rendering dynamic content directly as HTML within the widget. The example React widget uses `JSON.stringify` and direct text injection for the 'pizzaTopping', which are generally safe. The client environments (e.g., ChatGPT, MCP Inspector) are expected to sandbox these widgets.

Updated: 2025-12-23GitHub→

Medium Cost

jlcpcb-parts-mcp

by Adrie-coder

Sec9

Assists users in finding electronic components compatible with JLCPCB PCBA services by querying a local database.

Setup Requirements

⚠️Requires JLCPCB_DB_PATH environment variable to be set to a valid SQLite database file.
⚠️The SQLite database needs to be populated with JLCPCB parts data, likely from the 'jlcparts' project.
⚠️Requires Python 3.x and pip for dependency management.

Verified SafeView Analysis

The server uses parameterized SQL queries to prevent SQL injection vulnerabilities. The database path is loaded from an environment variable (JLCPCB_DB_PATH), which is good practice. Image fetching from external URLs relies on data stored in the database; trust in the data source (jlcparts project) is implied to prevent potential SSRF via malicious URLs.

Updated: 2026-01-19GitHub→

PreviousPage 425 of 760Next