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
karthik-s092003 icon

MY_FIRST_MCP_SERVER

by karthik-s092003

Sec4

An AI-powered system for generating software project structures and code in various languages, and converting code between languages.

Setup Requirements

  • ⚠️Requires API keys for OpenAI and/or Groq (paid services)
  • ⚠️Requires a Python environment with specific 'autogen' libraries and dependencies
  • ⚠️Relies on access to specific LLM models (e.g., gemini-2.0-flash-lite, gpt-oss-120b, llama-3.1-8b-instant)
Review RequiredView Analysis
The `create_folders` and `generate_and_write_files` tools directly interact with the filesystem (`os.makedirs`, `open`) using paths and content generated by LLM agents. This poses a critical risk of arbitrary file creation, modification, or deletion, including overwriting sensitive system files or injecting malicious code, if an adversarial prompt is used. There is no explicit sanitization or sandboxing of LLM-generated paths or file content. The `get_current_weather` tool also makes an external HTTP request with an LLM-provided parameter, which could be a minor vector.
Updated: 2026-01-04GitHub→
0
0
Low Cost
maguerrieri icon

expensify-heist-mcp

by maguerrieri

Sec7

Automates fetching and parsing Expensify CSV exports via Safari web automation on macOS, and can convert them to YNAB transaction format.

Setup Requirements

  • ⚠️Requires macOS
  • ⚠️Requires macOS System Settings → Privacy & Security → Automation permissions for the terminal/VS Code to control Safari
  • ⚠️Python 3.11+
  • ⚠️Safari browser must be installed and active
Verified SafeView Analysis
The server uses AppleScript to control Safari and inject JavaScript for web automation. The JavaScript snippets are hardcoded in `heist.py` to interact with expensify.com for login and report export. While this provides powerful control over the browser, the specific hardcoded actions appear benign. `subprocess.run` is used for AppleScript execution, which is a direct command execution. There are no clear indications of arbitrary code execution from untrusted input (e.g., `eval` on user-controlled data) or hardcoded secrets. Requires explicit macOS automation permissions, which acts as a user-controlled gate. A significant discrepancy exists between the provided source code (expensify-heist-mcp, using Safari) and the provided README (expensify-mail-mcp, describing Mail.app automation), which might cause confusion regarding its operational mechanism.
Updated: 2026-01-16GitHub→
0
0
High Cost
WouterArtsRecruitin icon

recruitin-mcp-servers

by WouterArtsRecruitin

Sec2

Automating Recruitin's complete sales and recruitment workflow, including labor market analysis, fee calculation, Pipedrive CRM integration, email/pitch generation, company insights, salary benchmarking, news gathering, and AI-powered document analysis for recruitment assessments.

Setup Requirements

  • ⚠️Numerous API Keys Required (Pipedrive, Brave Search, Notion, Typeform, Email SMTP credentials, Claude, Leonardo.ai, WhatsApp, Jotform, Figma). Many of these are hardcoded in provided scripts.
  • ⚠️Extensive Pipedrive CRM configuration (pipelines, stages, custom fields, native automations, webhooks) is required for full functionality, as detailed in multiple setup scripts and documentation.
  • ⚠️Requires Node.js 18.0.0+ runtime environment and `npm install` for various Node.js-based services.
  • ⚠️Requires manual cloud deployment (Vercel/Render) for specific webhook handlers and serverless functions to integrate with external services like Pipedrive webhooks.
Review RequiredView Analysis
CRITICAL: Multiple Pipedrive API tokens are hardcoded in test/setup scripts, and most critically, one is hardcoded as a fallback in a Vercel serverless function (`elite-email-composer-mcp/api/webhook/new-deal.js`) which is exposed to the internet. This grants full programmatic access to the CRM data. Several Notion API keys are also hardcoded in various Notion-related scripts. Brave API keys are hardcoded in some agents or provided as insecure fallbacks. Running this system without addressing these hardcoded secrets is extremely dangerous as it exposes sensitive API keys and potentially Pipedrive CRM data to unauthorized access.
Updated: 2026-01-17GitHub→
0
0
Medium Cost

mcp-server-library

by pramod-c-morn

Sec1

Provides a library or server implementation for a communication protocol, likely related to Minecraft, to enable custom server development or interaction.

Review RequiredView Analysis
Cannot perform a security audit as the source code was not provided for analysis. Without access to the code, it is impossible to check for 'eval', obfuscation, network risks, hardcoded secrets, or malicious patterns, making a reliable safety assessment impossible.
Updated: 2025-11-24GitHub→
0
0
Medium Cost
techmad220 icon

claude-code-mcp

by techmad220

Sec9

Exposes Claude Code CLI session history to Claude.ai or any MCP client for search and reference.

Setup Requirements

  • ⚠️Requires Claude Code CLI to be installed and used at least once to generate session history.
  • ⚠️Requires Rust toolchain (cargo) for compilation.
  • ⚠️Requires manual configuration in Claude Desktop's `claude_desktop_config.json`.
Verified SafeView Analysis
The server primarily operates locally, reading user session data from ~/.claude/. It uses SQLite with parameterized queries for indexing, preventing SQL injection. JSON-RPC tool calls are handled by predefined functions with argument validation, avoiding dynamic code execution. The SSE transport binds to 127.0.0.1 by default, limiting external exposure. The CORS policy is permissive ('Any') but is mitigated by the default local binding. No hardcoded secrets or obvious malicious patterns were found. File system access is scoped to user's .claude directory.
Updated: 2025-12-14GitHub→
0
0
Low Cost
jmcdice icon

superpower-mcp

by jmcdice

Sec9

Integrates the Superpowers skills library as tools for the Augment CLI, providing expert-crafted workflows to an AI coding assistant.

Setup Requirements

  • ⚠️Requires Node.js v18 or higher
  • ⚠️Requires Git installed
  • ⚠️Modifies user's Augment CLI configuration file (~/.augment/settings.json) during installation, with a backup created.
  • ⚠️Clones the Superpowers repository to ~/.augment/superpowers.
Verified SafeView Analysis
The server uses `StdioServerTransport`, meaning it communicates via standard input/output with the Augment CLI, not over a network, which minimizes network attack surfaces. The `install.sh` script performs file system operations (cloning a Git repository, creating directories, installing npm dependencies) and modifies the user's Augment settings JSON, but it includes backups and uses embedded Python for safer JSON manipulation. It also relies on a public, well-known repository for skills. No explicit `eval` or hardcoded secrets were found in the provided source.
Updated: 2025-11-27GitHub→
0
0
Low Cost
Imaginationfinagler591 icon

mcp-hello-world

by Imaginationfinagler591

Sec9

A minimal MCP server in Kotlin demonstrating the Model Context Protocol by exposing a simple 'greet' tool to clients via standard I/O.

Setup Requirements

  • ⚠️Requires Java Runtime Environment 8 or newer.
  • ⚠️Requires Gradle for building and running (if not using a pre-built JAR).
  • ⚠️Requires an MCP client (e.g., the provided HelloWorldClient or MCP Inspector) capable of communicating over Standard I/O.
Verified SafeView Analysis
The server implements a simple 'greet' tool with no external dependencies or dangerous operations, processing input locally. Communication is via standard I/O, typically for local process interaction. No 'eval' or hardcoded secrets were found in the server's code. The client (HelloWorldClient.kt) uses ProcessBuilder to launch the server JAR provided as a command-line argument; if the client were configured to accept untrusted input for this path, it could lead to arbitrary code execution, but this risk is on the client side, not within the server's logic itself.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
subsetsio icon

subsets-mcp-server

by subsetsio

Sec7

Provides a Model Context Protocol server for AI assistants to query and manage statistical datasets, either remotely from Subsets.io or locally via DuckDB.

Setup Requirements

  • ⚠️Requires a Subsets.io API Key (potentially paid) for remote mode functionality.
  • ⚠️Requires 'uv' (Python package installer) to be installed on the system.
  • ⚠️Requires Python 3.11 or higher.
  • ⚠️Initial setup involves downloading a 100MB+ semantic search embedding model and a keyword search index, requiring disk space and internet access.
Verified SafeView Analysis
The server's `execute_sql_query` tool directly executes arbitrary SQL queries provided as a string. While designed for AI-generated queries, this pattern can be a SQL injection vulnerability if the query parameter is not sanitized and comes from an untrusted source, potentially exposing or manipulating loaded datasets. The server relies on fetching and running code directly from a GitHub repository via `uvx`, introducing a supply chain risk if the repository were to be compromised. API keys and R2 credentials are read from files or environment variables, which is good practice.
Updated: 2026-01-07GitHub→
0
0
Medium Cost
jonathan-vella icon
Sec9

The Azure Pricing MCP server provides AI agents with real-time Azure retail pricing information for cost estimation, comparison, and SKU discovery in infrastructure planning workflows.

Setup Requirements

  • ⚠️Requires Python 3.10+ (if not using Docker)
  • ⚠️Requires internet access to the Azure Retail Prices API (prices.azure.com)
  • ⚠️Requires installation of Python dependencies (`pip install -e .[dev]` or `pip install -r requirements.txt`)
Verified SafeView Analysis
The MCP server interacts with a public Azure pricing API. While no direct 'eval' or obfuscation is found, parameters passed to the Azure Retail Prices API's '$filter' are directly inserted into the OData query string. This assumes the Azure API sanitizes inputs robustly. However, as it's a read-only public API, the impact of a theoretical injection would be minimal (e.g., fetching incorrect data, not arbitrary code execution). No hardcoded credentials for external services are found.
Updated: 2026-01-19GitHub→
0
0
Low Cost
jordimassaguerpla icon

mcp-server-uyuni-docs

by jordimassaguerpla

Sec10

This project provides documentation for an MCP Server integrated with the Uyuni system.

Verified SafeView Analysis
The provided 'SOURCE CODE' consists solely of a README.md file containing only the project title. There is no executable code to audit for security risks like eval, obfuscation, network issues, or hardcoded secrets. The README.md itself presents no security vulnerabilities.
Updated: 2025-12-05GitHub→
0
0
Low Cost
ofrausto3 icon

MCP_Server_Project

by ofrausto3

Sec8

A foundational FastAPI server for integrating and managing Spark and PostgreSQL data operations and ETL pipelines.

Setup Requirements

  • ⚠️Requires a Python environment with dependencies installed from `requirements.txt` (e.g., FastAPI, uvicorn, pydantic-settings, asyncpg, pyspark).
  • ⚠️For actual database interaction, a running PostgreSQL server accessible at the configured host/port is needed.
  • ⚠️For actual Spark operations, a Spark environment is required (defaults to `local[*]`, which needs Spark binaries/dependencies).
  • ⚠️Expects a `.env` file for configuration, though default values are provided in `config.py`.
Verified SafeView Analysis
No 'eval' or obfuscation found. Environment variables are used for sensitive configurations (e.g., Postgres credentials, Spark settings) via `.env` file, which is good practice. The current 'tools' implementations are stubs and do not yet interact with real external systems, minimizing immediate data exposure risk. The `uvicorn` server is configured with `reload=True`, which is suitable for development but should be disabled in a production environment.
Updated: 2025-11-21GitHub→
0
0
High Cost
acessosshmika icon

mcp-supabase-server

by acessosshmika

Sec7

Provides a Model Context Protocol (MCP) server for a sales agent, integrating with Supabase for data storage, Vertex AI for embeddings, and Cohere for re-ranking sales arsenal information. It aims to enhance N8N integrations with advanced AI capabilities.

Setup Requirements

  • ⚠️Requires a Google Cloud Platform project with Vertex AI enabled and a service account key (GCLOUD_SERVICE_KEY) configured for authentication and access.
  • ⚠️Requires a Supabase project with specific tables (`arsenal_vendas`, `leads`) and a custom PostgreSQL RPC function (`buscar_arsenal_vetorial`) for vector search, implying a pgvector setup.
  • ⚠️Requires a Cohere API key for the re-ranking functionality, which is a paid service and will incur usage costs.
Verified SafeView Analysis
Sensitive credentials (Supabase service key, Google Cloud service account key, Cohere API key) are correctly loaded from environment variables, preventing hardcoding. The `GCLOUD_SERVICE_KEY` is JSON parsed directly from an environment variable, which is a common pattern for Google Cloud credentials but requires that the environment variable's content is securely managed and not exposed or tampered with. The server utilizes default `cors()` (allowing all origins) and binds to `0.0.0.0`, which means it is publicly accessible by default unless deployed within a restricted network or behind a firewall. The use of a Supabase `SERVICE_KEY` grants elevated privileges and necessitates strict access control in production environments.
Updated: 2025-12-12GitHub→
PreviousPage 432 of 760Next