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
High Cost
ajay-fitbit icon

MCP

by ajay-fitbit

Sec3

Integrates Microsoft SQL Server databases with AI assistants (OpenAI, Claude Desktop) for natural language querying, automated database operations, and test suite generation.

Setup Requirements

  • ⚠️Requires OpenAI API Key (Paid)
  • ⚠️Requires Microsoft ODBC Driver 17 for SQL Server (Windows/Linux)
  • ⚠️Requires Python 3.8 or higher
  • ⚠️Requires access to a Microsoft SQL Server database with appropriate permissions
Review RequiredView Analysis
CRITICAL VULNERABILITIES IDENTIFIED: 1. SQL Injection Vulnerabilities: - In `server.py:DatabaseService.execute_query`, parameter *keys* are directly substituted into the SQL query via `query.replace(f"@{key}", "?")`. If a malicious AI prompt or user input controls the `key` name, SQL injection is possible (e.g., `@{key}; DROP TABLE Users; --`). - Similarly, in `server.py:DatabaseService.execute_stored_procedure`, parameter *names* are concatenated into the `EXEC` statement (`@{key} = ?`). A malicious agent controlling `key` names could inject SQL. 2. Local File Inclusion (LFI): In `server.py:DatabaseService.generate_query_from_template`, arbitrary `template_file` names are used to read files from the file system. A malicious actor could specify paths to sensitive files (e.g., `/etc/passwd`, `.env` files, or other configuration) leading to information disclosure. 3. Insecure Communication: `httpx.Client(verify=False)` is explicitly used in `openai_client.py`, `server_client.py`, `direct_no_mcp.py`, `direct_database.py`, and `stored_proc_explorer.py` for OpenAI API communication. While intended as a workaround for corporate SSL environments, this bypasses critical SSL certificate verification, making communication vulnerable to Man-in-the-Middle (MITM) attacks and compromising the confidentiality and integrity of API keys and potentially sensitive query contexts. 4. Subprocess Execution: Multiple launcher scripts use `subprocess.run` to execute other Python scripts. While currently controlled within the repository, this pattern could introduce risks if arguments or script paths become controllable by untrusted input, especially in conjunction with the identified injection vulnerabilities. RECOMMENDATIONS: The identified vulnerabilities are critical. Running this code in a production environment or with untrusted input is NOT recommended without significant security hardening.
Updated: 2025-11-19GitHub→
0
0
High Cost
hecatonheyrus icon

pdf-mcp-server

by hecatonheyrus

Sec3

Extracting textual content and generating summaries from PDF documents using Large Language Models (LLMs).

Setup Requirements

  • ⚠️Requires an OpenAI API Key (paid service) for LLM summarization, as it uses OpenAI LLM.
  • ⚠️Requires a Rust development environment (including Cargo) to build and run the server.
  • ⚠️The server is launched by specifying a single PDF file path as an argument, implying it processes one PDF per server instance.
Review RequiredView Analysis
Security analysis is severely limited as only the `README.md` file was provided, not the actual Rust source code (`.rs` files, `Cargo.toml`). Without access to the complete source code, it is impossible to audit for critical security risks such as potential `unsafe` blocks, malicious patterns, improper network handling, input sanitization vulnerabilities when interacting with LLMs, or hardcoded secrets. The server exposes HTTP/HTTPS endpoints for interaction, which can be a vector for attacks if not properly secured (e.g., authentication, authorization, TLS configuration). A full security audit would require the complete source code.
Updated: 2026-01-18GitHub→
0
0
Low Cost
OthmanAdi icon

CloudMcp

by OthmanAdi

Sec1

Provides a set of simple, executable tools (time, date, calculator, memory storage) via an MCP interface using Server-Sent Events (SSE).

Setup Requirements

  • ⚠️Requires Python and specific dependencies (e.g., uvicorn, starlette, mcp library). These are not explicitly listed in a requirements.txt but must be installed.
Review RequiredView Analysis
CRITICAL: The 'calculator' tool uses `eval()` on user-provided input ('expression'). This is a severe remote code execution vulnerability, allowing an attacker to execute arbitrary Python code on the server by injecting malicious expressions. The server is fundamentally unsafe to run in its current state without remediation.
Updated: 2025-11-21GitHub→
0
0
Medium Cost
iamfresh2 icon

mcp-server

by iamfresh2

Sec9

Manages and tracks personal or small business expenses by providing tools to add, list, and summarize financial entries.

Setup Requirements

  • ⚠️Requires Python 3.10 or higher.
  • ⚠️Requires local filesystem write access for the SQLite database in the system's temporary directory.
  • ⚠️Dependencies `aiosqlite` and `fastmcp` must be installed.
Verified SafeView Analysis
The application uses parameterized SQL queries ('?') which effectively prevents SQL injection vulnerabilities. It handles database initialization and operations asynchronously with `aiosqlite`, and includes error handling for database access issues (e.g., read-only mode). No 'eval' statements, hardcoded secrets, or obvious malicious patterns were found. The server binds to `0.0.0.0:8000`, which makes it accessible on all network interfaces; this is standard for local development but would require careful consideration and network security if exposed publicly.
Updated: 2025-11-19GitHub→
0
0
Medium Cost
consigcody94 icon

design-wand

by consigcody94

Sec8

Automate design workflows in Figma by providing AI-powered tools for file inspection, component management, asset export, style updates, and comment tracking.

Setup Requirements

  • ⚠️Requires generating a Figma Personal Access Token from your Figma account settings.
  • ⚠️Requires Node.js 18 or higher to run.
  • ⚠️Configuration for Claude Desktop requires specifying the absolute path to the server's executable file.
Verified SafeView Analysis
The server communicates with the Figma API using a user-provided personal access token. This token grants access to files the user can view. The server itself does not use 'eval' or obvious obfuscation. Input parameters are passed to the Figma API; potential vulnerabilities would mostly reside within the Figma API's handling of these inputs or how the client handles and stores the Figma token. The server handles and serializes JSON responses, which could reveal verbose error messages from the Figma API, but this is not a code execution risk. The README advises users to never commit tokens and to rotate them regularly.
Updated: 2025-11-25GitHub→
0
0
Medium Cost
ka1222te icon

growi-mcp

by ka1222te

Sec8

Facilitates interaction with Growi wiki instances through an MCP server, enabling AI tools to manage pages, attachments, and user information.

Setup Requirements

  • ⚠️Requires Growi API token, domain, and API version configured via environment variables (GROWI_DOMAIN, GROWI_API_TOKEN, GROWI_API_VERSION).
  • ⚠️Requires 'uv' or 'uvx' package manager for installation and execution.
  • ⚠️Some Growi servers may require a 'connect.sid' session cookie (GROWI_CONNECT_SID) for the 'download_attachment' function, which can be challenging to obtain.
  • ⚠️User registration via 'register_user' may require administrator approval on the Growi server depending on its settings.
Verified SafeView Analysis
The server uses environment variables for sensitive API keys (GROWI_API_TOKEN) and the Growi domain, which is good practice. No 'eval' or obvious malicious patterns are found in the provided source code. The primary security risk involves the permissions granted to the Growi API token and the potential for AI agents (especially with 'Auto-approve' enabled) to inadvertently expose private information or modify pages, as explicitly warned in the README. The 'download_attachment' function might require a 'connect.sid' session cookie for some Growi server configurations, which adds a dependency on browser-level authentication that needs careful handling.
Updated: 2025-12-01GitHub→
0
0
Low Cost
glebmish icon
Sec9

A Model Context Protocol (MCP) server that provides AI agents (specifically Claude Code) with tools for discovering OpenRewrite refactoring recipes and retrieving their documentation through semantic search.

Setup Requirements

  • ⚠️Docker and Docker Compose required (for the PostgreSQL database).
  • ⚠️Java 21 required (for the OpenRewrite documentation generator).
  • ⚠️Python 3.8+ required (for the MCP server implementation).
  • ⚠️Requires ~3GB disk space for generator artifacts, generated docs, and embedding models.
Verified SafeView Analysis
The server utilizes parameterized queries to prevent SQL injection and loads database credentials from environment variables, which are good security practices. It leverages a standard SentenceTransformer model for embeddings. The `startup.sh` script manages a local PostgreSQL Docker container in a controlled environment. While the broader workflow involves executing OpenRewrite recipes (which could theoretically perform arbitrary code changes), the MCP server itself is designed to provide search and documentation, and the `validate-recipe.sh` script uses trusted local init scripts to execute Gradle, mitigating direct malicious code injection through the recipe YAMLs. No 'eval' or obfuscation found.
Updated: 2026-01-02GitHub→
0
0
Low Cost
loveyou001 icon

github-mcp-server

by loveyou001

Sec10

This repository provides a GitHub Skills course that teaches users how to create a static website or blog using GitHub Pages and Jekyll.

Setup Requirements

  • ⚠️Requires a GitHub account to start the course and host the generated site.
  • ⚠️Private GitHub Pages repositories consume GitHub Actions minutes, which may incur costs depending on usage.
  • ⚠️Basic understanding of Git concepts (branches, commits, pull requests) is a prerequisite.
Verified SafeView Analysis
The provided content is a README file for a tutorial, not an executable server application. It contains no runnable code with security vulnerabilities like 'eval', obfuscation, or network risks. The only 'code-like' snippet is client-side JavaScript within an HTML comment, used to construct a GitHub URL, which poses no server-side security risk.
Updated: 2026-01-16GitHub→
0
0
Medium Cost
EdgeworthHitbox icon

dwr-mcp-server

by EdgeworthHitbox

Sec3

This server acts as a Model Context Protocol (MCP) interface, enabling AI models to retrieve various water-related data and administrative calls from the Colorado Division of Water Resources (DWR) REST API.

Setup Requirements

  • ⚠️Requires `npm run build` to compile TypeScript before running.
  • ⚠️No built-in authentication for HTTP/SSE mode; exposed public endpoints are accessible by anyone.
  • ⚠️The generic `query_dwr_api` tool allows access to any DWR API endpoint via the server, which could be misused due to lack of authentication.
Review RequiredView Analysis
The server explicitly states in its `SECURITY.md` and through code analysis that it has no authentication, making it unsafe for public or production use. Any client connecting to its `/sse` endpoint or `/api/*` REST endpoints will have full, unauthenticated access. Furthermore, the `query_dwr_api` tool allows arbitrary DWR API endpoints and parameters to be queried via the server's base URL, which, without authentication, could lead to abuse or unintended resource consumption on the DWR API.
Updated: 2025-11-19GitHub→
0
0
Medium Cost
infil00p icon

DriveBC_MCP

by infil00p

Sec9

Provides AI assistants with real-time BC highway conditions, road closures, and weather alerts via the Open511-DriveBC API.

Setup Requirements

  • ⚠️Requires Node.js 18 or higher to run due to ES module and fetch API usage.
  • ⚠️Designed to be used with MCP-compatible AI clients (e.g., Claude Desktop, LM Studio) via stdio transport, not as a standalone web service.
Verified SafeView Analysis
The server architecture is straightforward, using standard TypeScript practices and the Model Context Protocol SDK. It interacts with a public, unauthenticated API (Open511-DriveBC) and includes basic error handling, timeouts, and retries for API calls. Input validation for tool arguments is defined via JSON schema and further handled within the tool functions (e.g., region name validation, highway normalization). There are no detected uses of 'eval', direct child process execution without sanitization, or hardcoded sensitive credentials. The in-memory cache implementation is safe. Logging cache hits/misses to stderr is not a security flaw, but rather a minor operational detail. Overall, the codebase appears robust and secure for its intended purpose.
Updated: 2025-12-14GitHub→
0
0
Low Cost
sumikko-Charlotte icon

mcp-time-server

by sumikko-Charlotte

Sec9

A minimal Model Context Protocol (MCP) server that exposes a 'get_current_time' tool, returning the current system time.

Verified SafeView Analysis
The server operates via standard I/O (stdio_server), limiting direct network exposure unless explicitly piped. Tool calls are strictly validated against 'get_current_time', preventing arbitrary command execution. No 'eval', hardcoded secrets, or suspicious external calls were identified.
Updated: 2025-12-13GitHub→
0
0
Low Cost
flyfox666 icon
Sec7

Deploy a remote Model Context Protocol (MCP) server on Cloudflare Workers, providing an unauthenticated calculator tool for AI clients like Cloudflare AI Playground or Claude Desktop.

Setup Requirements

  • ⚠️Requires a Cloudflare account for deployment.
  • ⚠️Requires Node.js and npm/npx installed locally for development.
  • ⚠️The server is explicitly 'Authless', meaning no authentication is required for client connections. This is suitable for a demo but not for sensitive applications.
Verified SafeView Analysis
The server is explicitly designed to be 'authless', meaning it does not require authentication to access its tools. While this is stated in the README and acceptable for the simple calculator tool demonstrated, it is a significant security risk if deployed with more powerful or sensitive tools. The code itself does not contain obvious vulnerabilities like 'eval' or hardcoded secrets, and inputs for the calculator are schema-validated. However, the 'authless' nature means anyone can use the deployed API.
Updated: 2025-11-29GitHub→
PreviousPage 651 of 760Next