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)

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→
32
1
High Cost
hyperpolymath icon

claude-firefox-mcp

by hyperpolymath

Sec3

An MCP (Model Context Protocol) server that enables Claude Code CLI to control the Firefox browser for automation tasks.

Setup Requirements

  • ⚠️Requires Deno runtime to be installed.
  • ⚠️Requires manual loading and configuration of the Firefox browser extension via `about:debugging`.
  • ⚠️Requires Claude Code CLI for client-side interaction and specific configuration in `~/.claude/settings.json`.
Review RequiredView Analysis
The server and browser extension allow arbitrary JavaScript execution on any loaded webpage via the `execute_js` tool and through the `read_page` implementation that leverages `browser.tabs.executeScript` and `eval`. This grants full control over the browser's context to the Claude Code CLI, posing a significant risk if the CLI itself or the communication channel is compromised. The browser extension also requires broad `<all_urls>` permissions. While the Deno server binds to `localhost`, limiting direct external network attacks, local exploits or compromised client processes could lead to severe consequences, including data theft or malicious activity within the browser.
Updated: 2026-01-17GitHub→
32
2
Medium Cost
GhostTypes icon

pokemon-go-mcp

by GhostTypes

Sec9

A Model Context Protocol (MCP) server providing real-time Pokemon Go data (events, raids, research, eggs, Team Rocket, promo codes) via an API for integration with other tools.

Setup Requirements

  • ⚠️Python 3.10 or higher is required.
  • ⚠️Requires initial data scraping to populate the 'data/' directory. Without this, tools will return 'no data' messages. Example command: `python -c "from pogo_scraper.scraper import LeekDuckScraper; import asyncio; scraper = LeekDuckScraper(); asyncio.run(scraper.scrape_all())"`.
  • ⚠️A Docker image must be built before running via Docker: `docker build -t pogo-mcp-server .`.
  • ⚠️An MCP client (e.g., Claude Desktop, n8n, VS Code with MCP extension) is necessary to interact with the server's exposed tools.
Verified SafeView Analysis
The server primarily reads from local JSON data files populated by an integrated scraper. The scraper makes HTTP requests to leekduck.com, a trusted source, and parses HTML using BeautifulSoup. There are no identified instances of 'eval', obfuscation, hardcoded secrets, or malicious command execution from untrusted input within the server's core logic. The `data/` directory where scraped information is stored should be secured appropriately.
Updated: 2026-01-19GitHub→
32
20
Medium Cost
schemacrawler icon
Sec9

Provides an AI-powered MCP server for database schema analysis, SQL generation, and database interaction.

Setup Requirements

  • ⚠️Requires Docker and Docker Compose to run the server.
  • ⚠️Primarily designed for use with Visual Studio Code's MCP client for interaction.
  • ⚠️External database connections (e.g., MySQL, PostgreSQL) require setting `SCHCRWLR_DATABASE_USER` and `SCHCRWLR_DATABASE_PASSWORD` environment variables, typically in a `.env` file.
Verified SafeView Analysis
The provided configuration files, primarily Docker Compose definitions, do not contain hardcoded secrets or obvious malicious patterns. Database credentials are explicitly handled as environment variables, which is a good security practice. The overall security profile largely depends on the underlying `schemacrawler/schemacrawler-ai` Docker image, which is not part of the provided source code for analysis.
Updated: 2026-01-04GitHub→
32
2
Medium Cost
emilioSp icon
Sec9

Provides an MCP server and CLI tool for accessing iRacing racing simulation data API for AI assistant integration and direct command-line use.

Setup Requirements

  • ⚠️Requires an active iRacing account.
  • ⚠️Mandatory environment variables (COOKIE_JAR, EMAIL, and either PASSWORD or API_KEY) must be configured.
  • ⚠️An initial 'npm run login' command is required to authenticate and generate the cookie file for the first time.
Verified SafeView Analysis
The server uses environment variables for sensitive data (EMAIL, PASSWORD/API_KEY, COOKIE_JAR), avoiding hardcoding. It handles authentication cookies and refreshes them upon expiry. API calls are directed to the official iRacing API endpoints. No 'eval' or malicious patterns were identified. The primary security consideration is the local storage of the authentication cookie file, which should be secured on the host system.
Updated: 2025-11-26GitHub→
32
2
High Cost
yiwenlu66 icon

mu-mcp

by yiwenlu66

Sec8

Enable chat with AI models via OpenRouter, facilitating multi-model conversations and persistent state for AI agents.

Setup Requirements

  • ⚠️Requires an OpenRouter API Key (paid service).
  • ⚠️Designed to integrate specifically with Claude Desktop as an MCP server, not a standalone application for direct user interaction.
  • ⚠️Requires `uv` (recommended) or a Python virtual environment and `pip` for dependency management.
Verified SafeView Analysis
The server does not use `eval` or exhibit obfuscation. Secrets are handled via environment variables. The primary potential risk involves file handling: the `chat` tool accepts absolute file paths (`files`, `images`) as arguments. While this is expected functionality for an agent to provide context, it implies trust in the calling MCP client (Claude Desktop) to provide valid, non-malicious, and appropriately sandboxed paths. Path traversal within the context of arbitrary user-provided file paths could lead to unintended file disclosure if the calling client were compromised, though the server itself doesn't actively sanitize these paths beyond basic file existence checks. Conversation storage uses UUIDs for file names, mitigating direct path traversal for its own storage.
Updated: 2026-01-19GitHub→
32
2
High Cost
kengbailey icon

webintel-mcp

by kengbailey

Sec7

The WebIntel MCP server provides intelligent web search and content retrieval tools, including general web search, video search, web content fetching with pagination, YouTube transcription, and Reddit post/comment fetching, designed for AI assistants.

Setup Requirements

  • ⚠️Requires a running SearxNG instance configured for JSON output.
  • ⚠️Requires a running Speech-to-Text (STT) service (e.g., OpenAI compatible API) for YouTube transcription, which may incur costs.
  • ⚠️Docker is highly recommended for deployment, as setup instructions are tailored for it.
Verified SafeView Analysis
The server uses `os.getenv` for sensitive configurations like API keys and hostnames, which is good practice. Input validation is applied to tool parameters. However, the `WebContentFetcher` falls back to `Jina Reader API (r.jina.ai)` for some fetching failures or specific content types (e.g., PDFs). This means user-provided URLs and their content might be processed by a third-party service, which could have privacy and data leakage implications. Additionally, `yt-dlp` is configured with `nocheckcertificate: True`, which reduces the security of outbound connections by disabling SSL certificate verification, making it susceptible to man-in-the-middle attacks.
Updated: 2025-12-08GitHub→
32
1
Medium Cost
Storm00212 icon

MCP-test

by Storm00212

Sec1

Provides a futuristic terminal-like interface for an MCP server to launch applications, perform RAG queries on class notes, and integrate with engineering software.

Setup Requirements

  • ⚠️Requires OpenAI API Key (Paid) for RAG functionality.
  • ⚠️Requires Chrome browser for web-based tools (open_chrome, open_github, open_youtube).
  • ⚠️Hardcoded Windows paths for Proteus, MATLAB, LTSpice, and Git Bash executables, limiting cross-platform compatibility and requiring specific software installations and versions.
Review RequiredView Analysis
CRITICAL security risks identified. The server heavily relies on `child_process.exec` to launch applications and execute commands (e.g., `open_proteus`, `open_matlab`, `execute_git_bash_command`). User-provided input for parameters like `filePath`, `args`, and `command` is directly concatenated into shell commands without apparent sanitization, making the server highly vulnerable to remote code execution (RCE) via command injection. Furthermore, the Electron frontend (electron-app/src/main.js) uses `nodeIntegration: true` and `contextIsolation: false`, which are deprecated and insecure Electron settings, giving the renderer process direct access to Node.js APIs and the ability to invoke highly privileged MCP tools with malicious arguments. This combination presents an extremely high security risk.
Updated: 2026-01-18GitHub→
PreviousPage 183 of 760Next