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)

32
1
High Cost
kfreiman icon

vibecheck

by kfreiman

Sec9

A server for intelligent analysis of CVs and job descriptions, facilitating recruitment by assessing document fit and generating interview questions.

Setup Requirements

  • ⚠️Docker is required for easy setup and running.
  • ⚠️Requires a local Ollama/LLM instance for AI functionality (e.g., analysis, interview question generation), which often benefits from GPU support.
  • ⚠️HTML document processing relies on Playwright, which can have system-level dependencies or be resource-intensive.
Verified SafeView Analysis
The project demonstrates a strong awareness of common security risks, with explicit checks for path traversal ('..') and null bytes ('\x00') in file paths. Document processing for PDFs uses `go-pdfium` (WebAssembly), providing a sandboxed environment. HTML processing uses a headless Playwright browser and `go-readability` to extract content, which isolates rendering from the main application logic and attempts to remove boilerplate. No obvious hardcoded secrets or malicious patterns were found in the provided code. The reliance on distroless base images and structured error handling, as stated in the README, further enhances security posture. The system is described as a 'portfolio/demo project,' implying it may not be fully battle-hardened for all real-world attack vectors, but the explicit measures taken are robust.
Updated: 2026-01-19GitHub→
32
2
Medium Cost
Cylera icon
Sec9

This MCP server connects an AI assistant to Cylera's IoT security platform to provide LLMs with read-only access to device inventory, threats, risks, and utilization data.

Setup Requirements

  • ⚠️Python 3.13 required (strict version)
  • ⚠️Requires 'uv' for package management
  • ⚠️Cylera Partner API credentials required (CYLERA_USERNAME, CYLERA_PASSWORD, CYLERA_BASE_URL)
  • ⚠️Requires Docker Desktop for 'easy button' installation and for running the official image
Verified SafeView Analysis
The server is explicitly designed to be read-only (GET operations only), significantly reducing risks of data vandalism or unauthorized manipulation if an agent is compromised. Credentials are managed via environment variables and not hardcoded. The development workflow incorporates security scanning tools like Bandit and pip-audit. No dangerous functions like 'eval' or obfuscation were found. The explicit acknowledgment of data exfiltration risk and focus on mitigation in the README is a strong positive.
Updated: 2026-01-16GitHub→
32
2
Medium Cost
john-walkoe icon

uspto_pfw_mcp

by john-walkoe

Sec9

Provides enhanced search, data retrieval, and secure document management for USPTO Patent File Wrapper data, enabling legal research, due diligence, and cross-MCP (FPD, PTAB, Citations, Pinecone) integration.

Setup Requirements

  • ⚠️Requires USPTO_API_KEY (paid API) for core functionality.
  • ⚠️MISTRAL_API_KEY (paid Mistral AI API) is optional but required for advanced OCR capabilities.
  • ⚠️Native secure storage for API keys relies on Windows DPAPI; falls back to environment variables on other OS.
  • ⚠️An internal FastAPI proxy server runs on localhost (default 8080), automatically started in the background.
Verified SafeView Analysis
The server demonstrates a high level of security consciousness. It employs IP whitelisting (localhost only), CORS restrictions, robust security headers, and request size limits for its internal proxy. API keys are managed via unified secure storage (Windows DPAPI or environment variables) and are never exposed in URLs. Inter-MCP communication uses HMAC-SHA256 tokens instead of raw keys. All logging is sanitized to prevent sensitive data exposure (API keys, tokens, IPs, emails). Rate limiting is implemented for external USPTO API calls. The `INTERNAL_AUTH_SECRET` can be auto-generated if not provided, which is acceptable for development but should be explicitly configured in production environments as per `SECURITY_GUIDELINES.md`.
Updated: 2026-01-19GitHub→
32
1
Low Cost
nakada-memo icon
Sec8

Converts Excel design documents and spreadsheets into high-fidelity images, preserving visual elements and print layouts, optimized for large documents and external consumption.

Setup Requirements

  • ⚠️Requires Windows Operating System
  • ⚠️Requires Microsoft Excel (2016 or later recommended) to be installed
  • ⚠️Requires manual post-installation step for `pywin32`: `python Scripts/pywin32_postinstall.py -install`
Verified SafeView Analysis
The server interacts with Microsoft Excel via `win32com`, which requires Excel to be installed and running on a Windows OS. This design means the server directly interfaces with a desktop application, which inherently carries risks if processing untrusted Excel files (e.g., macro execution). However, the code itself does not introduce obvious vulnerabilities like `eval`, network risks, or hardcoded secrets. File paths are handled for temporary file creation and specified output directories. The `excel_app.DisplayAlerts = False` setting reduces user interaction but could hide potential Excel warnings. The primary security consideration is ensuring trusted Excel files are processed by the agent.
Updated: 2025-11-25GitHub→
32
2
Medium Cost
Asmah-Acheampong icon

TradingViewMCPServer

by Asmah-Acheampong

Sec9

Serves as a multi-asset trading assistant for Claude Desktop, providing real-time market data, technical analysis, and Pine Script v6 development tools for Forex, Stocks, and Crypto.

Setup Requirements

  • ⚠️Requires an Alpha Vantage API Key (free tier available, but registration is needed).
  • ⚠️Requires Python 3.10 or higher due to MCP SDK dependencies.
  • ⚠️While not strictly mandatory, Docker is recommended for production deployment, which adds an initial setup step for users unfamiliar with Docker.
Verified SafeView Analysis
The server demonstrates good security practices: - API keys (ALPHA_VANTAGE_API_KEY) are loaded from environment variables and not hardcoded. - The Pine Script sandbox explicitly states it performs static analysis and simulation, not actual code execution on the local Python interpreter, mitigating 'eval' related risks. - Network requests are handled with timeouts, retry logic (exponential backoff), and rate limiting, reducing the risk of network-related failures or abuse. - No obvious malicious patterns or obfuscation were found. - Input validation is implemented for various parameters.
Updated: 2026-01-19GitHub→
32
2
Medium Cost
triepod-ai icon
Sec7

Provides a comprehensive automated assessment platform for Model Context Protocol (MCP) servers, analyzing their security, functionality, protocol compliance, documentation, and various extended metrics. It acts as an inspector and auditing tool for MCP server developers and maintainers.

Setup Requirements

  • ⚠️Requires a separate Model Context Protocol (MCP) server to be available and configured as the target for assessment.
  • ⚠️The MCP Proxy server (part of this Inspector) generates a unique session token on startup, which is required for authenticating client-side connections (unless `DANGEROUSLY_OMIT_AUTH` is explicitly set).
  • ⚠️Optional Claude Code integration (for semantic analysis of tools and documentation) requires an external `mcp-auditor` service to be running, which can incur LLM token costs and may require additional API key configuration.
  • ⚠️Remote MCP servers being assessed may require OAuth 2.0 configuration (client ID, secret, scope).
Review RequiredView Analysis
The server component (proxy) is designed to execute arbitrary commands (`spawn-rx`) and make network requests to target MCP servers as part of its core security assessment functionality. This implies a high level of privilege is required to run this tool. Authentication for the proxy uses a session token with `timingSafeEqual` for robustness. Filename sanitization is applied to assessment output files to prevent simple path traversal. The client-side code defines numerous advanced attack patterns, but these are for testing *other* MCP servers, not vulnerabilities within this project itself. The tool performs checks for hardcoded secrets and insecure configurations in *assessed* MCP servers. Running this tool requires trusting its capabilities to interact with your system and network.
Updated: 2026-01-19GitHub→
32
1
Medium Cost
rogermt icon

forgesyte

by rogermt

Sec8

ForgeSyte is a modular AI-vision Model Context Protocol (MCP) server for image analysis, processing, and orchestration of pluggable vision modules, designed for integration with clients like Gemini-CLI.

Setup Requirements

  • ⚠️Requires Python 3.10+ and the `uv` dependency manager.
  • ⚠️API keys (`FORGESYTE_ADMIN_KEY`, `FORGESYTE_USER_KEY`) need to be configured for authenticated access. Anonymous access is only permitted if no keys are configured.
  • ⚠️Node.js 18+ and `npm` are required for developing and running the optional React Web UI.
Verified SafeView Analysis
The server implements API key-based authentication with SHA256 hashing and role-based access control (RBAC). Pydantic models are extensively used for input/output validation, reducing common web vulnerabilities. Structured logging and comprehensive error handling are present. The dynamic plugin loading system relies on Python entry-points, meaning plugins must be `pip install`able packages. While this allows extensibility, the security of the overall system depends on the trustworthiness and security of loaded plugins. The `ImageAcquisitionService` fetches external URLs, which is a potential SSRF vector, but it uses `httpx` with timeouts and status checks to mitigate risks. No direct `eval` or intentional obfuscation was observed.
Updated: 2026-01-19GitHub→
32
1
Low Cost
lionelschiepers icon

StockQuotes.MCP

by lionelschiepers

Sec9

Bridges AI assistants with real-time and historical financial market data from Yahoo Finance via the Model Context Protocol.

Setup Requirements

  • ⚠️Requires Node.js 24.0.0 or higher (LTS).
  • ⚠️Requires outbound internet access to Yahoo Finance.
  • ⚠️Dependencies must be installed via `npm install`.
Verified SafeView Analysis
The server utilizes `helmet` for security headers and `express-rate-limit` to prevent abuse on its HTTP transport. Input validation is enforced using Zod schemas. No hardcoded secrets or direct use of `eval` or `child_process` detected. The HTTP endpoint `/mcp` correctly restricts methods to POST. No authentication/authorization is implemented, which is typical for an MCP tool server expected to run in a controlled environment or behind an orchestrator.
Updated: 2026-01-19GitHub→
32
2
Medium Cost
aws-samples icon

sample-ecs-mcp-server

by aws-samples

Sec8

Deployment of an Agentic AI architecture on AWS Fargate using Amazon ECS, connecting to multiple Model Context Protocol (MCP) servers for tool execution.

Setup Requirements

  • ⚠️Requires Node.js (version >= 20)
  • ⚠️Requires Docker
  • ⚠️Requires AWS CLI configured
  • ⚠️Requires Bedrock Model Access to Anthropic Claude Sonnet 4
Verified SafeView Analysis
The architecture uses AWS Secrets Manager for the API key, and service-to-service communication employs Service Connect for private networking. IAM policies for Bedrock model invocation and S3 bucket listing, while using a wildcard resource (`*`), are explicitly justified for the sample's functionality. The Application Load Balancer is internet-facing, allowing HTTP access from any IPv4 address, which is noted but expected for a public API endpoint.
Updated: 2025-11-26GitHub→
32
2
Medium Cost
hangboss1761 icon
Sec7

Synchronize MCP services data from the official Cline Marketplace to a local SQLite database.

Setup Requirements

  • ⚠️Requires Node.js installed.
  • ⚠️Deletes `mcp_services.db` on every run, meaning all previous synchronized data will be lost and refetched.
  • ⚠️Requires continuous internet connectivity to `https://api.cline.bot` for synchronization.
Verified SafeView Analysis
The server synchronizes data from `https://api.cline.bot`. No `eval` or code obfuscation found. No obvious hardcoded sensitive API keys for its own operation. However, several dependencies in `pnpm-lock.yaml` are marked as deprecated or are older versions (e.g., `axios@0.21.4`, `sqlite3@5.1.7`, `rimraf@3.0.2`, `glob@7.2.3`), which could harbor unpatched vulnerabilities. The application deletes the local `mcp_services.db` file on every run before recreating it.
Updated: 2026-01-19GitHub→
32
2
High Cost
jonmmease icon
Sec3

Provide Rust debugging capabilities through LLDB's Python API, enabling clients to control debugging sessions, breakpoints, watchpoints, and execution flow.

Setup Requirements

  • ⚠️Requires Rust toolchain (cargo, rustc) to build and debug Rust targets.
  • ⚠️Requires Python 3.10-3.12 and the `lldb-python` package to be installed.
  • ⚠️On macOS, LLDB may require developer tools or specific code signing configurations to function correctly, potentially leading to permission errors.
Verified SafeView Analysis
The server uses LLDB's Python API functions like `frame.EvaluateExpression()` and `debugger.GetCommandInterpreter().HandleCommand()` which execute user-provided expressions or debugger commands directly within the debugged process's context. If the MCP server receives untrusted input, this could allow for arbitrary code execution or process manipulation in the debugging target. This is an inherent risk for a powerful debugger; therefore, the server requires deployment in a fully trusted environment with trusted clients. No hardcoded secrets or malicious patterns were identified in the source code.
Updated: 2025-11-28GitHub→
32
2
Low Cost
MateusTorquato icon

mcp-server-manager

by MateusTorquato

Sec9

Manage MCP servers for multiple AI clients through a unified CLI/TUI gateway.

Setup Requirements

  • ⚠️Requires Node.js >= 18 to run.
  • ⚠️The daemon requires an available port (default 8850); port conflicts will prevent startup.
  • ⚠️Certain shell configurations (e.g., zsh) are specifically handled, indicating potential compatibility nuances with other shells.
Verified SafeView Analysis
The project demonstrates good security practices, especially concerning user input validation for server commands (`validateCommand` checks for dangerous shell metacharacters). Sensitive data like API keys and OAuth client secrets are explicitly redacted for display. OAuth tokens are stored in the user's config directory, not hardcoded. Daemon operations involving system commands (e.g., `launchctl`, `systemctl`) appear to construct commands and paths carefully, mitigating shell injection risks. Network services are primarily local (HTTP gateway, OAuth callback server) on `localhost`, minimizing external attack surface.
Updated: 2026-01-12GitHub→
PreviousPage 172 of 713Next