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
ericskiff icon

mbta-mcp

by ericskiff

Sec9

Provides real-time MBTA Worcester Line train schedules as a Model Context Protocol (MCP) server for AI assistants.

Setup Requirements

  • ⚠️Requires a free MBTA API Key obtained from api-v3.mbta.com.
  • ⚠️The MCP server authentication (`MCP_API_KEY`) is optional; if not configured, the server will allow anonymous access, which may pose a security risk if deployed publicly without additional access controls.
  • ⚠️The project is built with Next.js and primarily designed for deployment on Vercel, requiring familiarity with Vercel's environment variable and serverless function management.
Verified SafeView Analysis
The server correctly utilizes environment variables for both the MBTA API Key and an optional MCP authentication key. Input validation for the 'get_worcester_departures' tool is robustly implemented using Zod schemas, mitigating common injection and data validation risks. Network requests to the external MBTA API are performed using `fetch` with controlled, parameterized inputs, preventing arbitrary code execution. The authentication mechanism for the MCP server itself (`MCP_API_KEY`) relies on a bearer token via the Authorization header; while optional, its implementation is standard when enabled. If the `MCP_API_KEY` is not set, the server allows anonymous access, which is a deployment configuration risk if not intended, but not a code vulnerability. No 'eval', direct command execution via user input, or obvious malicious patterns were found in the provided source code.
Updated: 2026-01-19GitHub→
0
0
Low Cost

llm-rag-mcp-example

by giuliolibrando

Sec8

Deploys an on-premise AI platform utilizing RAG and LLMs to manage IT infrastructure by indexing data from Redmine and Wiki.js.

Setup Requirements

  • ⚠️Requires Docker and Docker Compose for deployment.
  • ⚠️Manual LLM model pulling inside the 'ollama1' container is necessary after initial setup.
  • ⚠️Requires pre-existing Redmine and Wiki.js instances with configured API access for data ingestion.
Verified SafeView Analysis
The project emphasizes local deployment with Nginx Proxy Manager for external access and TLS termination. The README explicitly provides critical security advice regarding changing default credentials, protecting the NPM dashboard, and exposing only necessary ports. Adhering to these recommendations is crucial for maintaining a secure environment.
Updated: 2025-11-17GitHub→
0
0
Medium Cost
vaibhavdashoraIU icon

Trilix-Atlassian-MCP-Server

by vaibhavdashoraIU

Sec9

The Trilix Atlassian MCP Server enables AI assistants (like ChatGPT or Claude) to interact with multiple Atlassian workspaces (Confluence, Jira) simultaneously through a standardized Model Context Protocol (MCP). It acts as a trusted intermediary, providing normalized, validated, and secure data access.

Setup Requirements

  • ⚠️Requires Go 1.22+ to build and run.
  • ⚠️Requires Docker and Docker Compose for local infrastructure setup (PostgreSQL and RabbitMQ).
  • ⚠️The `TwistyGo` library must be manually cloned locally and referenced via a `replace` directive in `go.mod`, instead of standard `go get` dependency management.
  • ⚠️A 32-character `API_KEY_ENCRYPTION_KEY` must be securely generated and set in the environment for API token encryption.
  • ⚠️For user authentication, a Clerk.dev account and its `CLERK_SECRET_KEY` and `CLERK_PUBLISHABLE_KEY` are required, or the server will run in an unauthenticated development mode.
Verified SafeView Analysis
The system employs AES-256-GCM encryption for API tokens at rest in PostgreSQL, preventing sensitive data exposure. User authentication is handled via Clerk JWT, and a service token allows secure bot integration with user impersonation. The architecture explicitly avoids direct HTTP communication between microservices, opting for RabbitMQ, which enhances security. Documentation clearly outlines the necessity of replacing example secrets with strong, unique values. There are no indications of code injection vulnerabilities like 'eval' or other malicious patterns. The overall design prioritizes secure handling of credentials and user data.
Updated: 2026-01-13GitHub→
0
0
Low Cost
priyankawadle icon

Python_MCP

by priyankawadle

Sec9

Serves as a local backend for an AI agent (Claude) to manage employee leave requests, allowing the AI to check balances, apply for leave, and view history.

Setup Requirements

  • ⚠️Requires the Claude Desktop application to be installed to interact with the server.
  • ⚠️Relies on `uv` (Ultrafast Python package installer and resolver) for dependency management and execution, which might require an additional installation step if not already present.
  • ⚠️The `mcp` CLI tool is specifically used for installing and running the server, indicating a framework-specific setup process.
Verified SafeView Analysis
The source code uses standard Python operations (dictionary lookups, string formatting) and does not contain `eval`, `exec`, or direct shell command executions. Inputs are used as data (dictionary keys/values) and are not treated as executable code. There are no hardcoded secrets or sensitive credentials visible. The server's data is in-memory and non-persistent, reducing long-term data breach risks within the server itself.
Updated: 2025-12-16GitHub→
0
0
Low Cost
BridgerB icon
Sec8

Manages Minecraft servers remotely using the Minecraft Server Management Protocol (MCP) via an MCP server.

Setup Requirements

  • ⚠️Requires Deno 2.0 or higher.
  • ⚠️Requires a Minecraft Server 1.21.9+ with its management server enabled and configured in 'server.properties'.
  • ⚠️Uses Deno's unstable WebSocketStream API, requiring the '--unstable-net' flag, which may be subject to future breaking changes.
  • ⚠️Requires manual creation and editing of 'config.json' from 'config.json.example' with server details and secrets.
Verified SafeView Analysis
No 'eval' or obfuscation was found in the provided source code. The server communicates with its client (e.g., Claude Code) over stdio, which is secure. Connections to Minecraft servers use WebSockets with a bearer token from 'config.json'. Users must ensure 'config.json' containing these secrets is properly secured on their system. The use of the Deno '--unstable-net' flag is noted but not a direct security vulnerability.
Updated: 2025-11-22GitHub→
0
0
Low Cost
dsblank icon

gramps-ez-mcp

by dsblank

Sec8

Provides an MCP server to enable AI assistants to query and interact with Gramps genealogy family tree data using natural language.

Setup Requirements

  • ⚠️Requires Gramps genealogy software to be installed and an existing Gramps family tree database.
  • ⚠️Requires `pip install gramps` (or Gramps installed system-wide).
  • ⚠️The provided database name must exactly match an existing Gramps database tree (case-sensitive).
Verified SafeView Analysis
The server exposes direct database operations (e.g., `open_database`, `get_raw_person_data`) via external inputs (`dbname`, tool arguments). The `open_database(dbname, force_unlock=True)` call in `session.py` could be a vulnerability if `dbname` is not sufficiently validated by the Gramps library, potentially allowing path traversal or unauthorized database access. The direct mapping of tool names to Python functions and argument passing means the security heavily relies on the robustness of the underlying Gramps functions against potentially malicious inputs from the AI assistant or direct HTTP requests to the SSE endpoint. No obvious hardcoded secrets or direct `eval` calls.
Updated: 2025-12-05GitHub→
0
0
Low Cost
Ramakrishna-Gedala icon

mcp_emi_server

by Ramakrishna-Gedala

Sec9

This server exposes loan Equated Monthly Installment (EMI) calculation endpoints to MCP clients, acting as a proxy to a backend REST API.

Setup Requirements

  • ⚠️Requires Python 3.10+ and 'uv' package manager for environment management and execution.
  • ⚠️A separate EMI calculation REST API backend must be running and accessible at the `EMI_API_BASE_URL` (defaults to http://localhost:8000/api) for the server to function.
Verified SafeView Analysis
The server functions as a lightweight proxy and does not appear to contain malicious patterns, 'eval', or hardcoded sensitive secrets. It relies on an external, configurable backend for actual calculations, meaning its overall security is also dependent on the security of the proxied API. Input sanitization is implicitly handled by the backend API it calls.
Updated: 2025-11-27GitHub→
0
0
Medium Cost
bunnyf icon

pcb-mcp

by bunnyf

Sec9

Facilitates AI-assisted PCB design workflow by providing an MCP (Model Context Protocol) server for KiCad 9.x.

Setup Requirements

  • ⚠️Requires a Virtual Private Server (VPS) with Ubuntu 22.04+ or Debian 12+.
  • ⚠️Requires KiCad 9.0.6+, Python 3.10+, Java 17+ (for FreeRouting), and xvfb (for headless rendering) installed on the VPS.
  • ⚠️Requires an MCP client (e.g., Claude Code with MCP support) and SSH access configured to the VPS.
Verified SafeView Analysis
The server demonstrates strong security practices for its operations. It explicitly implements path validation (`is_safe_path`, `validate_file_path`) to prevent directory traversal attacks and restricts file reading to designated project and task directories. Shell injection is prevented by using `shlex.quote()` when constructing commands for background tasks (e.g., auto-routing). The server does not use `shell=True` for `subprocess.run` by default and does not contain obvious hardcoded secrets. It relies on SSH for secure communication, as it interacts via stdin/stdout rather than directly opening network ports. The main residual risk would stem from vulnerabilities in the underlying KiCad CLI, pcbnew API, or FreeRouting, but the server itself aims to interface with these tools securely.
Updated: 2026-01-17GitHub→
0
0
Medium Cost
Homalos icon

homalos-mcp

by Homalos

Sec8

A complete futures data analysis MCP service providing K-line data, technical indicators (KD, MACD), real-time updates via WebSocket, and historical data persistence in SQLite.

Setup Requirements

  • ⚠️Requires Node.js 20+ runtime environment.
  • ⚠️Relies on an external API (Sina Finance) for real-time data, which may have rate limits or availability issues.
  • ⚠️Requires `npm install` and `npm run build` before `npm start`.
Verified SafeView Analysis
The server fetches data from a third-party JSONP API (Sina Finance) and uses string manipulation followed by JSON.parse, which is safer than direct 'eval'. Futures contract symbols are concatenated into the API URL, which has a limited attack surface given the typical format of such symbols. WebSocket messages are also JSON parsed, and actions are restricted to known types ('subscribe', 'unsubscribe', 'ping'), minimizing arbitrary code execution risks. No apparent hardcoded secrets or malicious patterns.
Updated: 2025-11-27GitHub→
0
0
Low Cost
Khushil-21 icon

demo-mcp-server

by Khushil-21

Sec9

A simple Model Context Protocol (MCP) server providing utility tools for dice rolling and number addition for MCP-compatible clients like Claude Desktop.

Setup Requirements

  • ⚠️Requires Python >= 3.11
  • ⚠️Requires UV package manager for installation and running
  • ⚠️Manual path configuration for UV may be needed for Claude Desktop integration if UV is not in system PATH.
Verified SafeView Analysis
The server exposes two simple, non-destructive tools: 'roll_dice' and 'add_numbers'. There are no indications of 'eval' usage, hardcoded secrets, or complex logic that could be exploited. The `mcp.run` binds to `0.0.0.0` for remote accessibility, which is standard for a server but requires a secure deployment environment.
Updated: 2026-01-17GitHub→
0
0
Medium Cost
jrainbolt icon

McpServer

by jrainbolt

Sec9

Provides a REST API endpoint to proxy requests to OpenAI's chat models, serving as a backend for AI-powered chat functionalities.

Setup Requirements

  • ⚠️Requires an OpenAI API Key (Paid usage may apply).
  • ⚠️Requires Java Development Kit (JDK 17+) to build and run.
  • ⚠️Requires Maven or Gradle to build the Spring Boot application.
Verified SafeView Analysis
The application securely loads the OpenAI API key from environment variables or application properties, avoiding hardcoding. It uses WebClient with configured timeouts, retries for rate limits (429) and server errors (5xx), and comprehensive error logging, enhancing resilience and preventing unhandled API failures. No 'eval' or malicious patterns were found. The primary security risk involves proper management of the OpenAI API key outside the application.
Updated: 2025-11-24GitHub→
0
0
Medium Cost
BACH-AI-Tools icon

tavily-mcp

by BACH-AI-Tools

Sec9

Provides an MCP server to enable AI models with real-time web search, data extraction, website mapping, and crawling capabilities via the Tavily API.

Setup Requirements

  • ⚠️Requires a Tavily API key (can be obtained from tavily.com).
  • ⚠️Node.js v20 or higher is required.
  • ⚠️Potential for high token cost if `include_raw_content` is enabled or extensive crawl/extraction results are returned.
Verified SafeView Analysis
The server uses `process.env.TAVILY_API_KEY` for API key management, preventing hardcoded secrets. It relies on `@modelcontextprotocol/sdk` for schema validation of tool inputs, which helps mitigate injection risks. Network requests are handled by `axios`, a standard and secure HTTP client. No direct `eval` or shell command execution from user input was identified. Error handling for API calls is present. The server's primary function is to proxy requests to a third-party API (Tavily), limiting direct processing of arbitrary user input beyond defined tool schemas.
Updated: 2025-12-04GitHub→
PreviousPage 333 of 760Next