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
mohanchandrass icon

QA-MCP-Server

by mohanchandrass

Sec9

A knowledge-powered Q&A and action bot that uses the Model Context Protocol (MCP) to provide configurable responses, deterministic intent resolution, and controlled action triggering, suitable for enterprise support.

Setup Requirements

  • ⚠️Requires Python 3.11+
  • ⚠️Requires Docker (for the recommended server deployment method)
  • ⚠️Requires a Gemini API Key (must be set as GEMINI_API_KEY environment variable)
Verified SafeView Analysis
The system employs a strong security posture by strictly limiting the LLM's role to language generation and decoupling it from critical decision-making (intent resolution, action triggering, escalation). All core logic is deterministic and configuration-driven. Configuration files are loaded using `yaml.safe_load`, and API keys are retrieved from environment variables. No 'eval' or direct code injection vulnerabilities were found. The server binds to `0.0.0.0:8000`, a standard practice for Dockerized services, but requires proper network segmentation/firewalling in production.
Updated: 2026-01-17GitHub→
0
0
High Cost
developer-zht icon

mcp-servers

by developer-zht

Sec9

Provides AI assistants with the ability to summarize video and audio content, specifically Bilibili videos.

Setup Requirements

  • ⚠️Requires Node.js >= 18.0.0 and pnpm >= 8.0.0.
  • ⚠️Requires API keys for chosen LLM provider (e.g., OpenAI or Gemini), which are typically paid services.
  • ⚠️May require a Bilibili session token (AUTHORIZATION_BILIBILI_SESSION_KEY) for accessing certain video subtitles, potentially involving manual cookie extraction.
  • ⚠️Configuration relies on a `.env` file, which must be created and populated manually from `.env.example`.
Verified SafeView Analysis
Sensitive API keys (for LLMs and Bilibili sessions) are correctly managed via environment variables, reducing the risk of hardcoded secrets. External API calls are directed to trusted services (Bilibili, OpenAI, Gemini). The use of `JSON.parse` in a script is for processing trusted API stream data, not arbitrary user input. The project adheres to a monorepo structure with good practices for dependency management and TypeScript, which helps maintain code quality and prevent common vulnerabilities.
Updated: 2025-12-31GitHub→
0
0
Medium Cost
a-fleury icon

journal-me

by a-fleury

Sec3

An AI-powered conversational journaling application that helps users track life events, summarize activities, and get reminders by interacting with an LLM.

Setup Requirements

  • ⚠️Requires Java SDK 21 and sbt
  • ⚠️Requires Docker and Docker Compose for database setup
  • ⚠️Requires Cursor IDE for full MCP integration
Verified SafeView Analysis
CRITICAL: The application is explicitly designed to receive user credentials (email, password) and session tokens directly within LLM prompts (Model Context Protocol). This method is noted as a 'playful twist' for a 'school project' and 'Not recommended for production use'. However, this design fundamentally leaks sensitive information to the LLM and its provider, making it highly insecure for any real-world use case. Session tokens are also passed in the prompt, further increasing the risk of exposure.
Updated: 2025-12-15GitHub→
0
0
Low Cost
chatkausik icon

mcpserverexample

by chatkausik

Sec9

Provides a Model Context Protocol (MCP) server that exposes a simple 'add' tool for numerical operations.

Setup Requirements

  • ⚠️Requires Python 3.13 or higher (pyproject.toml specifies >=3.13, while README states 3.8+).
  • ⚠️Requires the 'uv' package manager for installation and local development.
  • ⚠️Installation and running directly from a Git URL using uvx requires an active internet connection.
Verified SafeView Analysis
The provided source code is minimal and implements a basic 'add' function. No 'eval', 'exec', shell command injection, hardcoded secrets, or obfuscation were found. The security score reflects the simplicity of the logic and the absence of obvious vulnerabilities in the presented code. The underlying 'FastMCP' framework's security is assumed but not audited here.
Updated: 2025-12-02GitHub→
0
0
Low Cost
chrisllontop icon

mcp-toolkit

by chrisllontop

Sec7

A desktop application for organizing and securely managing Model Context Protocol (MCP) servers and their configurations for various AI tools across different projects.

Setup Requirements

  • ⚠️Requires OS-native keychain access (macOS Keychain, Windows Credential Manager, Linux Secret Service).
  • ⚠️For development, requires a Rust toolchain, Node.js (v20+), and pnpm.
  • ⚠️When running tests, the development server (`pnpm dev`) must be running separately.
Verified SafeView Analysis
Uses OS-native keychain for storing the master encryption key, which is a strong security practice (AES-256-GCM encryption). A `MCP_TEST_MODE` environment variable can bypass keychain interaction and use a deterministic test key, explicitly marked for non-production use. The application's core functionality involves executing user-defined `command` and `args` for `Binary` and `Docker` MCP types, relying on user trust in configured MCPs. The frontend `tauri.conf.json` sets `csp: null`, which removes Content Security Policy, potentially increasing risk if an XSS vulnerability were to exist in the UI.
Updated: 2025-12-11GitHub→
0
0
Medium Cost
czabriskie icon

mcp-server

by czabriskie

Sec7

A generic, extensible Model Context Protocol (MCP) server providing weather and time tools, with an optional web interface for interactive testing with AWS Bedrock Claude models.

Setup Requirements

  • ⚠️Requires an AWS Account and configured AWS Bedrock access (paid services).
  • ⚠️Anthropic Claude models, necessary for MCP tool support, require filling out a 'Use Case form' in the AWS Bedrock console and typically have a ~15 minute approval time.
  • ⚠️Requires Python 3.10 or newer.
Verified SafeView Analysis
The `web_app` component is explicitly marked as 'not production-ready' in its README and source code, featuring broad CORS (`allow_origins='*'`), binding to all interfaces (`0.0.0.0`), and lacking authentication/authorization. This poses significant security risks if deployed in a production environment. The `TimeTools` module uses `http://ip-api.com/json` for geolocation, which is unencrypted HTTP, although the data transferred is not highly sensitive. The server is designed to run locally via stdio, limiting direct network exposure for the core components. AWS credentials must be configured securely via environment variables or AWS CLI, not hardcoded. Client IP injection into tools is handled with some sanitization (skipping private IPs) but does not prevent arbitrary public IP queries.
Updated: 2025-11-22GitHub→
0
0
High Cost
mrvivekkr icon
Sec6

Generates AI-powered Markdown and HTML documentation for GitHub repositories by analyzing their structure, open issues, pull requests, and branches.

Setup Requirements

  • ⚠️Requires OpenAI API Key (Paid)
  • ⚠️Requires GitHub Personal Access Token (with 'repo' and 'user' scopes)
  • ⚠️Requires manual building and copying of custom Go binaries (`github-mcp-server` and `mcpcurl`)
Verified SafeView Analysis
The system executes external Go binaries (`mcpcurl` and `github-mcp-server`) using `subprocess.Popen` with arguments constructed from user input (GitHub owner, repo, path). While `subprocess` is used with a list of arguments (mitigating shell injection), the security relies heavily on the robustness of these external binaries against argument injection or malicious input parsing. Additionally, a hardcoded Django `SECRET_KEY` and `DEBUG = True` are present, which are unsuitable for production environments.
Updated: 2025-11-24GitHub→
0
0
High Cost
krishnab0841 icon

Multi_MCP_AI_Agent

by krishnab0841

Sec3

The Multi-MCP Intelligent Assistant enables natural language interaction with GitHub, Perplexity, Calendar, and Gmail by integrating multiple Model Context Protocol (MCP) servers.

Setup Requirements

  • ⚠️Requires API keys/credentials for integrated services (GitHub, Perplexity, Calendar, Gmail).
  • ⚠️Requires an API key for the underlying Large Language Model (LLM) (e.g., OpenAI, Anthropic, or other MCP-compatible LLMs), which may incur costs.
Review RequiredView Analysis
No source code for the server was provided for analysis, only the README.md file. Therefore, a comprehensive security audit of the server's implementation (e.g., checking for 'eval', hardcoded secrets, network risks, or malicious patterns) could not be performed. The score reflects the inability to assess the server's safety based on the provided information.
Updated: 2025-11-25GitHub→
0
0
Low Cost
betterhyq icon

honeycomb

by betterhyq

Sec8

A service configuration management platform based on Model Context Protocol (MCP), providing visual MCP service configuration and management.

Setup Requirements

  • ⚠️Requires Node.js >= 24.11.1 (or specific version from .nvmrc).
  • ⚠️Requires pnpm >= 10.25.0 (or specific version from package.json).
  • ⚠️Mandatory database initialization step: 'pnpm init-db' before first run.
  • ⚠️The Docker setup uses an internal JD.com pnpm registry (`http://registry.m.jd.com/`), which might not be accessible externally. Users might need to configure pnpm to use a public registry.
Verified SafeView Analysis
The project uses `dotenv` for environment variable management and has a specific CORS origin configured, which are good security practices. The `ToolModel` and database schema include a `callback` field of type `string` intended to store JavaScript code. However, in the provided server-side `mcp.ts` file, the actual MCP server `registerTool` method uses a fixed anonymous function, and the `tool.callback` string from the database is currently unused for execution. This means there is no immediate Remote Code Execution (RCE) vulnerability from `eval`ing user-controlled code in the provided snippets. This is a critical distinction from an earlier assessment. Future implementations that activate this `callback` string for dynamic execution would introduce a severe RCE risk without robust sandboxing. Standard web application vulnerabilities (e.g., XSS, CSRF, insecure dependencies) may still apply but cannot be fully assessed from truncated code.
Updated: 2026-01-19GitHub→
0
0
Low Cost
arvindand icon

mcp-scaffold

by arvindand

Sec9

A Maven plugin that generates Spring AI Model Context Protocol (MCP) tools from Spring Data JPA repositories and Spring services, enabling AI assistants to interact with the application's data layer.

Setup Requirements

  • ⚠️Requires Java 21+ for compilation and runtime.
  • ⚠️Maven 3.9+ is required, Gradle projects are not currently supported.
  • ⚠️The target application must be a Spring Data JPA application.
  • ⚠️Requires adding `mcp-annotations` and `spring-ai-starter-mcp-server-webmvc` dependencies to the project's `pom.xml`.
Verified SafeView Analysis
The MCP Scaffold Maven plugin itself is safe, focusing on build-time code generation using JavaParser and JavaPoet. It does not involve runtime 'eval' or obfuscation. The generated MCP tools, by design, provide direct access to the application's data layer. The documentation explicitly highlights this, advising users to implement authentication, restrict write operations using `filter.exclude-methods` (a feature the `suggest` goal aids in), and limit exposure to read-only functionality for AI clients. The `ReadOnlyDetector` and `suggest` goal actively contribute to identifying and preventing exposure of dangerous write operations.
Updated: 2026-01-17GitHub→
0
0
Low Cost
fancier21 icon
Sec8

Deploys a remote Model Context Protocol (MCP) server on Cloudflare Workers to expose simple calculator and random number generation tools without requiring authentication.

Setup Requirements

  • ⚠️Requires Node.js and npm to install dependencies and the Cloudflare Wrangler CLI for deployment and local development.
  • ⚠️Connecting to local MCP clients (e.g., Claude Desktop) requires an additional 'mcp-remote' proxy.
  • ⚠️The 'authless' nature means that if sensitive tools were added, external authentication/authorization would need to be implemented separately.
Verified SafeView Analysis
The server is explicitly 'authless', meaning it exposes its tools publicly without authentication, which is stated as its intended design. For the provided simple calculator and random number tools, this is low risk. Input validation for tool arguments is performed using Zod schemas, which helps mitigate common injection vulnerabilities. External network requests are made to drand.cloudflare.com, a trusted randomness beacon. No hardcoded secrets or obvious malicious patterns were found.
Updated: 2025-11-19GitHub→
0
0
Medium Cost
dvy246 icon

Mcp-server

by dvy246

Sec8

A Streamlit and CLI-based chat interface leveraging the Model Context Protocol (MCP) to connect Google's Gemini LLM with various tools and servers for tool-augmented conversational AI.

Setup Requirements

  • ⚠️Requires Google Gemini API Key (Paid)
  • ⚠️Python 3.10+ required
  • ⚠️Requires `uv` to run the default local math server configuration
Verified SafeView Analysis
The core application code appears secure, utilizing `yaml.safe_load` for configuration and loading API keys from environment variables. There are no obvious 'eval' or malicious patterns. However, the Model Context Protocol (MCP) design inherently involves executing external processes (local or remote) as 'tools' configured by the user (e.g., `MATH_SERVER_PATH`, `MANIM_SERVER_PATH`, or the remote `expense` server). While this is the intended functionality, it means the overall security depends heavily on the trustworthiness and proper securing of these external tools/servers that the application interacts with. Users must ensure that the configured server paths point to safe executables/scripts.
Updated: 2025-11-20GitHub→
PreviousPage 336 of 760Next