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
Medium Cost
jcornell3 icon
Sec6

Provides a local development and testing environment for Model Context Protocol (MCP) servers, enabling integration with Claude Desktop and external APIs.

Setup Requirements

  • ⚠️Requires Docker and Docker Compose for the local development environment.
  • ⚠️Windows users must have WSL2 installed and properly configured for Docker integration and `wsl docker exec` commands.
  • ⚠️Many services require specific API keys (e.g., `MCP_API_KEY`), GitHub Personal Access Tokens (`GITHUB_PERSONAL_ACCESS_TOKEN`), and Google Cloud credentials for full functionality.
  • ⚠️The `santa-clara` server's Playwright-based scraper and `youtube-to-mp3` server's `yt-dlp` tool introduce external software dependencies that need proper installation and resource management.
Verified SafeView Analysis
The environment involves running multiple Docker containers, some of which require high privileges, such as the `real-test-bridge.js` which executes `docker exec` commands. This implies access to the Docker daemon socket, a significant security risk if the bridge or its container is compromised, potentially leading to root access on the host. The `youtube-to-mp3` server downloads content from external sources (YouTube), which carries inherent risks of malicious files or large file-based denial-of-service. Interaction with sensitive external APIs (GitHub, Google Drive) relies on securely configured API keys and personal access tokens (PATs); if not properly managed (e.g., using default API keys in production), these could be exposed. While individual server code generally avoids common injection vulnerabilities and uses bearer token authentication, the overall architecture is designed for a development environment and requires careful hardening for production use, especially regarding Docker socket access and external dependencies.
Updated: 2025-12-08GitHub→
0
0
Low Cost
mdownes icon
Sec9

An AI agent that leverages a microservice backend (MCP server) to use tools for fetching weather information and performing basic math operations.

Setup Requirements

  • ⚠️Requires OpenAI API Key (Paid) for the LLM.
  • ⚠️Requires running two separate Python processes concurrently (`mcp_server.py` and `main.py`) for the full system to function.
Verified SafeView Analysis
The system is generally safe. The `mcp_server.py` listens on `0.0.0.0` (all interfaces), which could expose the tool execution API if the machine has a public IP. However, the `agent.py` is configured to call `localhost:8000`, limiting internal exposure. The tools themselves (`weather_tool`, `math_tool`) have very restricted functionality (dummy data, simple math) and do not expose any dangerous system commands or file access, mitigating the risk of external abuse.
Updated: 2025-11-19GitHub→
0
0
High Cost
ScholarChen20 icon

travel_agent

by ScholarChen20

Sec3

An AI-driven smart travel planning assistant that leverages a multi-agent system, integrates with map services (Amap), provides social features, and includes administrative/monitoring tools.

Setup Requirements

  • ⚠️Requires an LLM API Key (e.g., OpenAI, DeepSeek, Alibaba Cloud Qwen) which can incur costs.
  • ⚠️Requires Amap API Key for mapping services, which may incur costs.
  • ⚠️Requires MySQL, MongoDB, and Redis databases to be running and accessible.
  • ⚠️Requires `uvx` and `amap-mcp-server` command-line tools, likely installed as part of the HelloAgents framework or globally.
  • ⚠️Backend requires Python 3.10+; Frontend requires Node.js 16+.
Review RequiredView Analysis
Critical security vulnerabilities found. The `AMAP_MAPS_API_KEY` is hardcoded in `backend/app/agents/trip_planner_agent.py` and `backend/app/services/amap_service.py`, which is a severe information leak. The CORS configuration in `backend/app/api/main.py` is overly permissive with `allow_origins=... + ['*']`, making it vulnerable to various cross-origin attacks in a production environment. Default sensitive values like `jwt_secret_key`, `mysql_password`, and `redis_password` are present in `config.py`, posing a risk if not explicitly overridden by environment variables. File upload handling includes basic type and size validation but a thorough review for more complex bypasses (e.g., polyglot files) would be prudent. The application uses ORM for database interactions, which generally protects against SQL/NoSQL injection, but overall security is significantly compromised by hardcoded keys and permissive CORS.
Updated: 2026-01-19GitHub→
0
0
Low Cost
jkf87 icon

weather-mcp

by jkf87

Sec9

Provides weather information for major Korean cities via an MCP server, integrating with AI assistants.

Setup Requirements

  • ⚠️Requires Python 3.10 or higher
Verified SafeView Analysis
The server uses `httpx` to make requests to the Open-Meteo API. City inputs are validated against a predefined list, preventing injection vulnerabilities. No hardcoded secrets or API keys are used (as per Open-Meteo's design). No `eval` or other dynamic code execution patterns were found. The use of a fixed external API (Open-Meteo) reduces the attack surface significantly. A score of 9 instead of 10 is given as any external API dependency carries a minimal inherent risk, though Open-Meteo is a well-known and generally reliable service.
Updated: 2025-11-30GitHub→
0
0
Medium Cost
Sec8

This application generates SQL queries from natural language prompts by fetching database schema context from an MCP server and utilizing an LLM.

Setup Requirements

  • ⚠️Requires OpenAI API Key (Paid service for GPT-4 access).
  • ⚠️Requires a separate MCP Server instance running, typically on port 8080, to provide database schema context.
  • ⚠️Requires Java 17 or higher.
Verified SafeView Analysis
The application acts as a client, relaying user prompts and fetched database schema context to an external LLM (OpenAI GPT-4). The primary security considerations include potential prompt injection vulnerabilities (common with LLM integrations) and the trust placed in the MCP server, as its provided schema context directly influences the generated SQL. The application itself does not execute the generated SQL, mitigating direct SQL injection risks within the client. OpenAI API key is handled securely via environment variables.
Updated: 2025-11-18GitHub→
0
0
Medium Cost
jphyqr icon
Sec9

An MCP server to register, verify, and audit AI prompts for security with HashBuilds, providing embeddable badges for websites.

Setup Requirements

  • ⚠️Requires Claude Code for integration
  • ⚠️Requires Node.js version 18 or higher
  • ⚠️Relies on external HashBuilds API for prompt scanning and verification
  • ⚠️Requires the absolute path to 'dist/index.js' in Claude Code configuration
Verified SafeView Analysis
The server acts as a thin client/proxy to the external HashBuilds API. It does not perform local code execution (e.g., no 'eval' or direct 'child_process' calls beyond `fetch`). The security of the prompts themselves, and the prompt scanning, relies entirely on the external HashBuilds API. No hardcoded secrets are evident, and the API base URL is configurable via an environment variable.
Updated: 2025-12-12GitHub→
0
0
Medium Cost
proxziima icon
Sec8

Enables LLM agents to autonomously explore Databricks Unity Catalog metadata, analyze data lineage, examine processing code, and execute SQL queries to answer data-related questions.

Setup Requirements

  • ⚠️Requires a Databricks workspace with Unity Catalog enabled and configured.
  • ⚠️Requires `DATABRICKS_HOST`, `DATABRICKS_TOKEN`, and `DATABRICKS_SQL_WAREHOUSE_ID` environment variables, demanding specific Databricks credentials and permissions.
  • ⚠️Python 3.10+ is required.
Verified SafeView Analysis
The server properly uses environment variables for Databricks credentials, explicitly warns against committing `.env` files, and recommends using service principals with minimal permissions. The `execute_sql_query` tool grants significant power to the LLM agent, requiring careful management of the associated Databricks token's scope to prevent unauthorized data access or modification. No `eval` or obfuscation found. Network calls are made to the Databricks API as expected for its functionality.
Updated: 2025-11-26GitHub→
0
0
Medium Cost

python-docs-mcp-server

by EzraFathurrahman

Sec8

A Python server application leveraging LiteLLM to process or serve documentation, potentially enabling LLM-powered features for multi-cloud platform documentation.

Setup Requirements

  • ⚠️Requires access to one or more LLM API keys (e.g., OpenAI, Anthropic, Google) as LiteLLM integrates with paid services, incurring usage costs.
  • ⚠️Lack of a `README.md` and `requirements.txt` file means users will need to manually infer and install dependencies, and configure the server based on `config.py` and `server.py` contents.
  • ⚠️Requires a suitable Python environment (version needs to be inferred) to run, with all necessary packages installed.
Verified SafeView Analysis
The server interacts with external LLM APIs via LiteLLM. Potential risks include API key exposure (if `config.py` is mishandled), unauthenticated access leading to billing abuse, and prompt injection if user inputs are not properly sanitized for LLM prompts. Without code, a precise assessment is difficult, but LiteLLM itself is a robust library. Assumes standard secure coding practices in `server.py` and proper network configuration.
Updated: 2025-11-17GitHub→
0
0
High Cost
nohuiam icon

skill-builder

by nohuiam

Sec8

Manages a centralized repository of AI skills (SKILL.md files), enabling their creation, validation, retrieval, matching, and usage tracking within a cognitive ecosystem.

Setup Requirements

  • ⚠️Requires build tools (e.g., C++ compiler) for 'better-sqlite3' if prebuilt binaries are unavailable for your specific Node.js version and operating system, which can be a common dependency issue.
  • ⚠️The 'src/disk-loader.ts' contains hardcoded skill directories (`/Users/macbook/Documents/claude_home/repo/claude-skills`, `/Users/macbook/Documents/claude_home/repo/bop/skills`) that will need to be manually created or adjusted for skills to load from disk on startup.
  • ⚠️Requires network ports 3029 (UDP for InterLock), 8029 (HTTP API), and 9029 (WebSocket) to be free and accessible on the host system.
Verified SafeView Analysis
The server employs input sanitization for skill names to prevent potential path traversal or injection issues. It uses a UDP-based InterLock mesh for internal communication, which includes a configurable signal whitelist to filter unexpected messages, providing a basic layer of defense. No 'eval' or blatant hardcoded secrets were identified within the provided code. The Linus security audit confirmed no critical issues and noted fixes for major/minor input validation concerns, which are now addressed.
Updated: 2026-01-17GitHub→
0
0
Medium Cost
versaprav1 icon

mcp-server-python

by versaprav1

Sec4

Provides Model Context Protocol (MCP) interfaces for interacting with PostgreSQL databases and an integration platform API covering Azure, SAP BTP, and SAP ABAP endpoints.

Setup Requirements

  • ⚠️Requires a PostgreSQL database instance to be running and accessible.
  • ⚠️Requires an '.env' file to be configured with database and API credentials.
  • ⚠️The Integration Platform API Server ('api_server.py') requires a separate backend REST API server ('rest_api_server.py') to be running on port 3000 to function, creating a multi-tier setup.
  • ⚠️Public exposure via ngrok requires manual installation and authentication of ngrok.
Review RequiredView Analysis
The 'postgres_server.py' includes an 'execute_query' tool that allows executing arbitrary SQL queries directly against the PostgreSQL database. While it supports parameterized queries for values, this function itself represents a significant SQL injection risk if not used with extremely careful input validation or strict access control, as it can execute DDL, DML, or administrative commands. The 'rest_api_server.py' uses f-strings for schema names in its SQL queries, but the 'schema' parameter is strictly validated against a whitelist ('dev', 'prod', 'test'), mitigating direct schema name injection. The use of Ngrok for public exposure is noted as optional but requires careful security measures (authentication, IP whitelisting) as advised in the README. No hardcoded secrets were found; sensitive data is loaded from .env files. Overall, the direct SQL execution capability in the PostgreSQL server is the primary critical security concern.
Updated: 2025-12-09GitHub→
0
0
Medium Cost
that0n3guy icon
Sec8

Enables AI assistants to interact with a Mealie recipe database, allowing for recipe search, creation, updates, and meal planning.

Setup Requirements

  • ⚠️Requires Node.js 18+ runtime environment.
  • ⚠️Requires a running Mealie instance accessible via network.
  • ⚠️Requires a valid Mealie API key for authentication.
Verified SafeView Analysis
The server correctly uses environment variables for `MEALIE_BASE_URL` and `MEALIE_API_KEY`, which is a good practice to avoid hardcoding secrets. It handles API requests, including image uploads via URL or base64 data, and includes error formatting to prevent raw internal error exposure. No `eval` or direct `child_process` execution was observed. The primary security considerations depend on the underlying Mealie instance's security and the secure management of the API key in the deployment environment.
Updated: 2025-12-13GitHub→
0
0
Medium Cost
sydasif icon
Sec7

Nornir MCP Server exposes network automation capabilities to Claude, enabling natural language interaction for monitoring and managing network infrastructure.

Setup Requirements

  • ⚠️Requires Python 3.11+.
  • ⚠️Requires network access to target devices via SSH with valid credentials.
  • ⚠️Requires manual configuration of Nornir inventory files (config.yaml, hosts.yaml, groups.yaml, defaults.yaml) with live device details and credentials.
Verified SafeView Analysis
The codebase includes several positive security features such as input validation with Pydantic models, configurable command blacklisting (`CommandValidator`) to prevent dangerous CLI commands, path traversal prevention for backup directories, and sensitive data sanitization in resource outputs (`_sanitize_dict`). However, a critical security risk lies in the provided example configuration files (`examples/conf/hosts.yaml`, `examples/conf/groups.yaml`, `examples/conf/defaults.yaml`) which contain hardcoded usernames and passwords (e.g., 'cisco', 'admin', 'secure_password'). The quick start guide encourages copying these examples, potentially leading users to deploy the server with insecure, plain-text credentials in a production environment. While `exec` is used in `resources.py` for dynamic function wrapping, its usage appears contained and controlled, injecting only trusted functions and parameters.
Updated: 2026-01-19GitHub→
PreviousPage 552 of 760Next