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
Low Cost
mmcdermott-ashleyne icon

fde_sql_mcp

by mmcdermott-ashleyne

Sec9

Provides an MCP server for programmatically interacting with on-prem SQL Server instances using Windows authentication, offering tools to enumerate database objects.

Setup Requirements

  • ⚠️Python 3.10+ required.
  • ⚠️ODBC Driver 17 or 18 for SQL Server must be installed on the host.
  • ⚠️The account running the server requires Windows authentication permissions to the target SQL Server instance.
  • ⚠️SQL Server host address must be configured via 'fde_sql_mcp.config.json' or 'SQL_SERVER_HOST' environment variable.
Verified SafeView Analysis
The server relies on Windows authentication (Trusted_Connection=yes) and explicitly externalizes SQL Server connection details to 'fde_sql_mcp.config.json' or environment variables, avoiding hardcoded secrets. All SQL queries used by the tools are fixed strings targeting system views ('sys.databases', 'sys.tables', etc.), which inherently mitigates SQL injection risks. The 'pyodbc' library is a standard and secure choice for database connectivity. 'TrustServerCertificate' is configurable. No 'eval' or obfuscation found.
Updated: 2026-01-16GitHub→
0
0
High Cost
Lewis-R-L icon

gpt-mcp-server

by Lewis-R-L

Sec2

An MCP server integrating with italki's API to provide language learning services, including teacher recommendations, language/country metadata, and personal calendar events, with an optional OAuth 2.0 authentication provider.

Setup Requirements

  • ⚠️Weak Password Hashing: The mock OAuth provider uses SHA256 for password hashing, which is insecure for production use.
  • ⚠️Access Token Logging: When `LOG_REQUEST_RESPONSE=true`, the full Authorization header (including access tokens) is logged to console, posing a severe security risk.
  • ⚠️Non-persistent Data in Serverless: The embedded NeDB database (`OAUTH_DB_PATH=/tmp/db` on Vercel) and in-memory MCP sessions are not persistent across serverless function invocations or server restarts, making the mock OAuth provider unsuitable for production.
Review RequiredView Analysis
CRITICAL ISSUES: 1. Weak Password Hashing: The mock OAuth provider (NeDBUsersStore) uses SHA256 for password hashing, which is not cryptographically secure for user credentials in production environments and is highly susceptible to brute-force and rainbow table attacks. The code acknowledges this is 'for demo purposes' but it's a severe vulnerability if used. 2. Access Token Logging: When `LOG_REQUEST_RESPONSE` environment variable is set to `true`, the full `Authorization` header (including sensitive access tokens) is explicitly logged in plaintext to the console (main.ts, `oauthHeaders.authorization` in access log middleware). This is a severe security risk as access tokens grant access to user data and should never be logged. 3. Wildcard CORS for Admin API: The admin API (admin-server.ts) uses `Access-Control-Allow-Origin: *` (wildcard CORS). While the local version runs on a separate port, the Vercel deployment integrates it into the main app. Exposing an administrative interface with such permissive CORS without strong authentication and origin checks is a high security risk. OTHER ISSUES: - Non-persistent Data in Serverless: The in-memory MCP sessions and the NeDB database (when `OAUTH_DB_PATH` points to `/tmp/db` on Vercel) are not persistent across serverless function invocations or server restarts, leading to data loss for OAuth clients, users, and sessions. This is a functional limitation, not a direct security flaw in the code logic itself, but impacts the reliability and security posture of the OAuth provider in a production serverless environment.
Updated: 2025-12-22GitHub→
0
0
Medium Cost
HenFaibishProjects icon

mcp-gitpower-server

by HenFaibishProjects

Sec9

This server acts as a Model Context Protocol (MCP) tool provider, allowing an agent to list GitHub repositories for a specified organization or user.

Setup Requirements

  • ⚠️Requires a GitHub Personal Access Token (GITHUB_TOKEN) with sufficient permissions (e.g., 'repo' scope for private repos, or public_repo for public ones).
  • ⚠️Requires a GitHub organization or username (GITHUB_ORG) to be specified.
  • ⚠️Requires Node.js (version 18 or higher is recommended based on dependencies).
Verified SafeView Analysis
The code correctly loads GitHub credentials from environment variables, avoiding hardcoded secrets. It utilizes a well-known SDK (@octokit/rest) for GitHub interactions. The application's core logic for handling tool calls appears straightforward and free from obvious malicious patterns or 'eval' usage. Input validation is handled by the MCP SDK's schemas (ListToolsRequestSchema, CallToolRequestSchema). The main security concern would be ensuring the GITHUB_TOKEN has appropriate, least-privilege scopes.
Updated: 2025-12-01GitHub→
0
0
Medium Cost
jordanlixu icon

eth-mcp-server

by jordanlixu

Sec9

Provides Ethereum-related tooling like balance queries, token price lookup, and Uniswap swap simulation for AI agents via a Model Context Protocol (MCP) interface.

Setup Requirements

  • ⚠️Requires numerous environment variables (`INFURA_URL`, `WALLET_ADDRESS`, `USDC`, `BTC`, `WETH`, `UNI`, `BETH`, `UNISWAP_V2_ROUTER`), which must be set to valid Sepolia testnet addresses or URLs.
  • ⚠️The `rmcp` Rust SDK is pulled directly from a GitHub `main` branch, which might introduce stability concerns compared to a released version and requires `git` tooling.
  • ⚠️Relies on an Ethereum RPC provider (e.g., Infura) which will incur costs based on usage.
Verified SafeView Analysis
The server explicitly states that it does NOT execute real transactions; swap functionality uses `eth_call` for safe simulation only, significantly reducing financial risk. There are no indications of 'eval', obfuscation, or hardcoded sensitive secrets. Environment variables are used for RPC URLs and contract addresses. The use of `.unwrap()` in some places can lead to panics on unexpected errors or invalid input, which is a robustness concern but not a direct security vulnerability in terms of arbitrary code execution.
Updated: 2025-11-19GitHub→
0
0
Low Cost
SunmaiLee icon
Sec8

Boilerplate for creating Model Context Protocol (MCP) servers with integrated AI capabilities (e.g., image generation, code review prompts) using TypeScript and Hugging Face.

Setup Requirements

  • ⚠️Requires a Hugging Face API token (`hfToken`) for the image generation tool.
  • ⚠️Requires `npm install` and `npm run build` (or `npx smithery build`) for setup and compilation from TypeScript to JavaScript.
  • ⚠️Built using the Smithery framework, which expects a `createServer` function as the default export in `src/index.ts`.
Verified SafeView Analysis
Hugging Face API token (hfToken) is explicitly required via configuration, preventing hardcoding. Input validation for tools is enforced using Zod schemas, which helps mitigate common input-related vulnerabilities. The server makes external API calls to Hugging Face for image generation, introducing a dependency on that service's security. The 'code_review' tool generates an LLM prompt but does not execute code itself; any execution risk lies with the client's LLM environment.
Updated: 2025-11-27GitHub→
0
0
Low Cost
tonyfruzza icon

vice-mcp-server

by tonyfruzza

Sec7

Provides AI assistants with debugging and control capabilities for the VICE C64 emulator.

Setup Requirements

  • ⚠️Requires Python 3.7+.
  • ⚠️Requires VICE C64 emulator running with the '-remotemonitor' flag enabled.
  • ⚠️Requires an MCP-compatible client (e.g., Claude Desktop) for interaction.
Verified SafeView Analysis
The server connects to a local VICE emulator's remote monitor interface, reducing network exposure. It includes a `send_monitor_command` tool which allows arbitrary commands to be sent to the emulator. While this grants significant control over the emulated environment, it does not directly expose the host system to arbitrary command execution. No 'eval' or obvious hardcoded secrets were found.
Updated: 2025-12-15GitHub→
0
0
Low Cost
prateekmulye icon

mcp-server-client-demo

by prateekmulye

Sec9

This project demonstrates the integration of custom Micro-Core Protocol (MCP) servers providing tools with a LangChain client.

Setup Requirements

  • ⚠️Requires OpenAI API Key (Paid)
  • ⚠️Requires Python 3.13 or newer
Verified SafeView Analysis
The provided code is simple and does not contain obvious security risks such as `eval`, `exec`, or hardcoded credentials. It properly uses `python-dotenv` for API keys. The exposed functions perform basic arithmetic and return a hardcoded string. Security would largely depend on the `mcp.server.fastmcp` framework's implementation, which is not fully visible.
Updated: 2025-12-01GitHub→
0
0
High Cost
Sec8

Real-time anti-cheat server utilizing AI (GPT-4, MiniMax M2) to detect cheating patterns and suspicious activity in games.

Setup Requirements

  • ⚠️Requires Node.js >= 18.0.0
  • ⚠️Requires Discord Bot Token, Application ID, and Owner ID for core functionality.
  • ⚠️OpenAI API Key required for GPT-4 features (paid service).
  • ⚠️MiniMax API Key required for MiniMax M2 features (paid service).
Verified SafeView Analysis
This anti-cheat server requires deep system access (memory, network, processes, registry) for its core functionality. The project states adherence to privacy principles (no permanent data storage, anonymous analysis, encrypted communication, no API keys in logs) to mitigate associated risks. No explicit `eval` or obfuscation detected, suggesting a relatively safe implementation given its purpose. However, running any anti-cheat system always implies granting high privileges to the software.
Updated: 2025-11-23GitHub→
0
0
Low Cost
azzizulhakim810 icon

Basic_Mcp_Server

by azzizulhakim810

Sec8

A Model Context Protocol (MCP) server providing basic utilities like arithmetic, conversational prompts, and Google Calendar integration for AI agent consumption.

Setup Requirements

  • ⚠️Requires a Google API Key (needs Google Cloud project setup).
  • ⚠️Requires a specific Google Calendar ID to access event data.
  • ⚠️Requires `npm install` to set up dependencies and `npm run build` to compile TypeScript.
Verified SafeView Analysis
The server correctly uses environment variables (`process.env`) for sensitive information like the Google API Key and Calendar ID, preventing hardcoding. No 'eval' or obfuscation is present. The primary security consideration is the secure management of these environment variables in the deployment environment. The naming `GOOGLE_PUBLIC_API_KEY` for an authentication credential could be misleading, as such keys are typically sensitive and should not be publicly exposed.
Updated: 2025-12-09GitHub→
0
0
Medium Cost
chef72max icon
Sec9

An AI-powered digital reading companion application that assists users with note-taking, highlighting, goal tracking, and provides AI-driven insights and summaries for various document types like PDFs and EPUBs.

Setup Requirements

  • ⚠️Requires Node.js 18+ to run the development server.
  • ⚠️AI features are currently placeholders ('TODO: Replace with actual AI API call') and require custom implementation and potential API key integration (e.g., OpenAI, Gemini) for full functionality, which would incur external costs.
Verified SafeView Analysis
The application primarily operates client-side, storing user data (books, notes, highlights) in IndexedDB and localStorage, which limits direct server-side vulnerabilities from the provided code. The AI features are currently placeholders, meaning actual AI API keys/integrations are not present and thus no hardcoded secrets or immediate network risks associated with AI calls exist in this version. File upload is simulated. Loading PDF worker from CDN (jsdelivr) is a common practice but relies on CDN integrity. Overall, the code appears to be clean and focuses on client-side functionality, making it relatively safe to run.
Updated: 2025-12-15GitHub→
0
0
Medium Cost
prem676 icon
Sec2

Provides a semantically enriched search experience over AWS Cloudscape Design System documentation, leveraging AI to understand and retrieve information effectively.

Setup Requirements

  • ⚠️Requires Python 3.7 or later, 4GB RAM, and 200MB disk space.
  • ⚠️The `SentenceTransformer` model ('Alibaba-NLP/gte-multilingual-base') is downloaded on the first run of the ingestion script, requiring internet access.
  • ⚠️The installation and run commands in the provided README (e.g., `pip install -r https://raw.githubusercontent.com/...zip`, `python https://raw.githubusercontent.com/...zip`) appear to be malformed, pointing to a `.zip` file within a 'docs/components/badge' path rather than a `requirements.txt` or a Python script. This would prevent successful setup and execution.
  • ⚠️The `trust_remote_code=True` setting for `SentenceTransformer` is a significant security concern for ML supply chain attacks.
Review RequiredView Analysis
The ingestion script (`ingest.py`) uses `trust_remote_code=True` when loading the `SentenceTransformer` model. This is a critical supply chain security risk, as it allows arbitrary code execution if the referenced model on Hugging Face or its repository is compromised. More severely, the `cloudscape_read_doc` function exposed as an MCP tool takes a `file_path` argument directly from what can be user input or an AI agent. There is no input validation or sanitization, making the server vulnerable to Path Traversal attacks, allowing an attacker to read arbitrary files on the system (e.g., `/etc/passwd`, sensitive configuration files).
Updated: 2026-01-19GitHub→
0
0
Medium Cost
Ujwalmahajan777 icon

test-mcp-server

by Ujwalmahajan777

Sec9

This server acts as an expense tracker, allowing users to add, list, and summarize personal financial expenses using a local SQLite database.

Setup Requirements

  • ⚠️Python 3.10+ required
Verified SafeView Analysis
SQL injection is effectively prevented by using parameterized queries for all database operations. The SQLite database is created in a temporary directory, mitigating risks related to persistent storage in sensitive locations. No 'eval' or obvious obfuscation is present. There are no hardcoded secrets or sensitive credentials. Error messages, while catching broad exceptions, provide a controlled message (`Database error: {str(e)}`) which is better than exposing full stack traces.
Updated: 2026-01-10GitHub→
PreviousPage 296 of 760Next