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
tamangmilan icon
Sec8

This server provides a simple API for tracking, listing, and summarizing personal expenses.

Setup Requirements

  • ⚠️Requires Python 3.13 or newer
Verified SafeView Analysis
The server uses parameterized SQL queries, which effectively prevents common SQL injection vulnerabilities. No 'eval', 'exec', or other potentially dangerous functions were identified. There are no apparent hardcoded secrets. The server is exposed via HTTP, but the provided code snippet does not include authentication or authorization mechanisms, which would be necessary for secure deployment in untrusted environments.
Updated: 2025-11-30GitHub→
0
0
Medium Cost
professordnyc icon

qanat-goose-mcp

by professordnyc

Sec8

A multimodal seller dashboard assistant for Square, integrating voice and gesture controls via a Goose MCP-UI extension.

Setup Requirements

  • ⚠️Requires a Square API Key for sandbox or production (a free developer account or paid account is needed).
  • ⚠️Requires an ElevenLabs API Key for voice features (ElevenLabs offers free and paid tiers).
  • ⚠️PyAudio is required for microphone input, which can be challenging to install on some operating systems. Voice features will be disabled if not present.
  • ⚠️Requires 'Goose Desktop' application to run as an MCP-UI extension.
Verified SafeView Analysis
The project uses environment variables for API keys and is structured to avoid common vulnerabilities like `eval` with user input. However, the demo scripts include hardcoded `demo_key` values as fallbacks, which could be a concern if these fallbacks were to be used in a production context. The main `env_loader` explicitly validates required keys. Inputs from the Goose MCP-UI are passed directly to service methods; robust input validation within services for production use is crucial.
Updated: 2025-11-22GitHub→
0
0
Medium Cost
Sec9

A Ruby implementation of the Model Context Protocol (MCP) server for QuickBooks Online integration, enabling CRUD and search operations for various QuickBooks entities.

Setup Requirements

  • ⚠️Requires Ruby 3.0 or higher
  • ⚠️Requires a QuickBooks Online developer account and app credentials (Client ID, Client Secret)
  • ⚠️Requires configuring `http://localhost:8000/callback` as a Redirect URI in the QuickBooks app settings
Verified SafeView Analysis
The server uses standard OAuth 2.0 for authentication and relies on environment variables for sensitive credentials (client ID, client secret, refresh token, realm ID). It can operate as an HTTP server locally, requiring proper configuration and securing of the host environment. The project leverages well-established gems (quickbooks-ruby, oauth2). No obvious signs of 'eval' or obfuscation were found. Running an HTTP server and handling financial data inherently requires careful security practices in deployment and credential management.
Updated: 2025-11-18GitHub→
0
0
Low Cost
JJAMAWAVE icon

mcp-relay-server

by JJAMAWAVE

Sec6

The MCP Relay Server acts as an intermediary, forwarding tool call requests from clients to a local agent and synchronizing tool definitions.

Setup Requirements

  • ⚠️Requires a 'tools/' directory containing Python modules with `@mcp_tool` decorated functions for tool definitions.
  • ⚠️Requires a separate 'Local Agent' application to connect via WebSocket to `/ws` to provide actual tool implementations; without it, tool calls will fail.
  • ⚠️Relies on `uvicorn` for execution (implicitly, as it's a FastAPI app).
Verified SafeView Analysis
The server uses an open CORS policy (`allow_origins=["*"]`) which allows requests from any domain. More critically, the WebSocket endpoint (`/ws`) used by the 'Local Agent' is unauthenticated. An attacker could potentially connect to this endpoint, register arbitrary tools, and intercept/relay tool calls if the server is exposed publicly without network-level security. The tool loading mechanism from a local 'tools/' directory is generally safe, assuming the server's host environment is trusted and not susceptible to arbitrary file writes.
Updated: 2025-11-22GitHub→
0
0
Medium Cost
yuvko24 icon
Sec9

Provides an API wrapper for accessing country information from an external REST API (restcountries.com), exposing tools to query country details by name, list by region, or search by language.

Setup Requirements

  • ⚠️Requires Python environment with FastAPI, FastMCP, httpx, and uvicorn installed.
  • ⚠️Relies on the external restcountries.com API availability and its data structure; changes to this API could break functionality.
Verified SafeView Analysis
The server uses `urllib.parse.quote` for user-provided input in URLs, which helps prevent URL injection. It uses `httpx` with a timeout, which is good practice for external API calls. No `eval` or hardcoded secrets are present. The primary external risk comes from the reliability and security of the `restcountries.com` API itself, which is beyond the scope of this code's direct control. The server binds to 0.0.0.0, making it externally accessible if deployed without proper firewalling, but this is standard for server applications.
Updated: 2025-11-23GitHub→
0
0
Low Cost
jules-tnk icon

ts-http-mcp

by jules-tnk

Sec8

An MCP server enabling AI agents to send HTTP requests with full control over methods, headers, query parameters, and request bodies.

Setup Requirements

  • ⚠️Requires Node.js environment
  • ⚠️Only supports stdio transport for MCP communication
Verified SafeView Analysis
The server is designed to proxy HTTP requests from AI agents. While the code includes robust input validation (Zod schema for parameters, URL format validation), deployment in environments with access to sensitive internal networks could pose an SSRF (Server-Side Request Forgery) risk if the AI agent's prompts are not sufficiently sandboxed or if the agent itself is compromised. No obvious malicious patterns, hardcoded secrets, or dangerous functions like 'eval' were found in the provided source code. It uses well-vetted libraries like Axios for HTTP requests and Zod for schema validation, enhancing internal security.
Updated: 2025-11-23GitHub→
0
0
High Cost
brandonlacoste9-tech icon

skills-integrate-mcp-with-copilot

by brandonlacoste9-tech

Sec7

This repository demonstrates how to integrate GitHub Copilot Agent Mode with Model Context Protocol (MCP) servers to practice agentic loops for various development stages including research, planning, implementation, testing, and updates.

Setup Requirements

  • ⚠️Requires GitHub Copilot subscription/access
  • ⚠️Likely requires API keys/authentication for external MCP services (GitHub, Figma, Playwright)
Review RequiredView Analysis
Insufficient source code provided for a comprehensive security audit. The provided `README.md` file does not contain any executable code or patterns like 'eval', obfuscation, or hardcoded secrets. A full audit would require inspecting the actual implementation files.
Updated: 2025-11-24GitHub→
0
0
High Cost
taskcrew icon

cua-mcp-server

by taskcrew

Sec9

Delegates desktop automation tasks to an autonomous vision-based AI agent controlling cloud-based virtual machine sandboxes.

Setup Requirements

  • ⚠️Requires CUA Cloud API Key (Paid for sandbox access)
  • ⚠️Requires Anthropic API Key (Paid for vision AI processing)
  • ⚠️Requires Vercel Pro plan (for 800s maximum function timeout)
Verified SafeView Analysis
The server implements robust input validation for sandbox names and task IDs to prevent injection and path traversal attacks. It includes specific URL validation (`isValidBlobUrl`) to prevent Server-Side Request Forgery (SSRF) when fetching task history from Vercel Blob storage. API keys are sourced from environment variables or request headers, not hardcoded. The agent architecture relies on an LLM generating structured tool calls, not direct shell commands from user input, significantly reducing direct execution risks. The primary security concern, as noted in the `AGENTS.md` and `README.md`, is the default wide-open CORS policy (`Access-Control-Allow-Origin: *`) for broad MCP compatibility, which might need tightening for production deployments with untrusted users, and the implication that the server's Anthropic API key is shared among authenticated CUA API key users.
Updated: 2025-12-17GitHub→
0
0
Low Cost
Mkhubaiib icon

mcp_calculator

by Mkhubaiib

Sec9

A beginner-friendly Model Context Protocol (MCP) server exposing a simple calculator tool (add, subtract, multiply, divide) over WebSocket using FastAPI.

Setup Requirements

  • ⚠️Requires Python 3.9 or higher.
  • ⚠️Dependencies must be installed via pip (e.g., 'pip install -e .').
Verified SafeView Analysis
The server demonstrates strong security practices for its scope: it uses Pydantic for robust inbound message validation, handles errors gracefully by returning structured error messages without exposing internal details, and explicitly catches division-by-zero errors. No 'eval' or similar dangerous functions are used. There are no hardcoded secrets, and dependencies are standard. The server primarily performs arithmetic, limiting the attack surface. A basic HTML demo page is exposed on `/demo`, which might typically be removed in a production environment, but is benign in this context.
Updated: 2025-11-27GitHub→
0
0
Medium Cost
horison-ai icon

outlook-mcp

by horison-ai

Sec9

Provides Microsoft Outlook/Graph API integration as an MCP server, enabling email management and calendar event scheduling.

Setup Requirements

  • ⚠️Requires a separate 'connectivity service' to be running and accessible, typically at http://localhost:8001, to obtain user credentials.
  • ⚠️Requires Python 3.11 or higher.
  • ⚠️Depends on a 'shared' Python module for `credential_cache`, implying it's part of a larger project structure and not a standalone utility.
Verified SafeView Analysis
The server securely handles credentials by fetching them from an external 'connectivity service' via environment variables, avoiding hardcoded secrets. It uses `httpx` for API calls and raises exceptions on HTTP errors. No 'eval' or obvious malicious patterns were found.
Updated: 2025-11-22GitHub→
0
0
Medium Cost
akbad icon
Sec9

Provides AI-enhanced code search capabilities by integrating with a Sourcegraph instance via the Model Context Protocol (MCP).

Setup Requirements

  • ⚠️Requires access to a Sourcegraph instance (self-hosted or sourcegraph.com).
  • ⚠️Requires Python 3.10+.
  • ⚠️Requires configuration of `SRC_ENDPOINT` environment variable.
Verified SafeView Analysis
Uses environment variables for sensitive data like API tokens. Employs `yaml.safe_load` for parsing prompt configurations, mitigating YAML deserialization risks. Performs network requests to user-configured Sourcegraph endpoints. Server binds to `0.0.0.0` (all interfaces) by default, standard for a server application.
Updated: 2026-01-19GitHub→
0
0
Medium Cost

go.model-orchestrator

by TypeTerrors

Sec8

Orchestrates multiple LLM agent wrappers and tool servers via mDNS, presenting a single OpenAI-compatible API to clients.

Setup Requirements

  • ⚠️Requires an OpenAI-compatible LLM service (e.g., Ollama, which it defaults to) to be running and accessible for all agents.
  • ⚠️Relies on mDNS for service discovery, which may require specific network configuration or permissions to function correctly across components.
  • ⚠️Requires a Go runtime environment with dependencies for manual execution, or Docker/Docker Compose for containerized deployment.
Verified SafeView Analysis
Relies on mDNS for discovery within a local network and exposes HTTP endpoints. The primary orchestrator API (OpenAI compatible) and child agent/tool server APIs should be run in a secured, isolated network environment (e.g., Docker internal network) and not exposed directly to the public internet without proper security measures. Tools are limited to HTTP verbs, reducing arbitrary code execution risk but still requiring careful consideration of what URLs and data they can access.
Updated: 2025-11-17GitHub→
PreviousPage 419 of 760Next