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
trangpl193 icon

strange-mcp-core

by trangpl193

Sec9

Provides a shared foundation and framework for building Model Context Protocol (MCP) servers that expose AI-callable tools over HTTP.

Setup Requirements

  • ⚠️Requires Node.js version 18 or higher.
  • ⚠️API Key is required if `ApiKeyAuth` middleware is enabled, typically loaded from `process.env.API_KEY`.
  • ⚠️This package is a framework; a developer must write a custom server implementation using `McpServer` and register tools for it to be functional.
Verified SafeView Analysis
The framework itself is well-engineered for security, providing robust JSON-RPC handling, API key authentication, and structured error responses (e.g., preventing stack trace leaks from operational errors). Input body size is limited to prevent DoS. CORS is enabled with `Access-Control-Allow-Origin: *`, which is permissive but common for APIs. The primary security consideration for a deployed server built with this framework will lie in the implementation of the `ToolDefinition.handler` functions, as these are arbitrary code executed in response to network requests. The framework provides the mechanisms but cannot guarantee the security of custom tool implementations.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
marc-shade icon
Sec7

Provides comprehensive security scanning capabilities for targets and cluster nodes using Nuclei, integrated with an Agentic System for autonomous vulnerability management.

Setup Requirements

  • ⚠️Requires Nuclei vulnerability scanner binary to be installed and accessible via NUCLEI_BIN environment variable or default path (~/go/bin/nuclei).
  • ⚠️The 'sentence-transformers' library (if available) for anomaly detection and prioritization features may download a large AI model ('all-MiniLM-L6-v2') on first use, requiring internet access and local storage.
  • ⚠️Relies on the AGENTIC_SYSTEM_PATH environment variable for default configuration paths.
Verified SafeView Analysis
The server executes external commands via `asyncio.create_subprocess_exec`, which is generally safer than shell-based execution as arguments are passed as a list. However, parameters like `target` and `severity` are directly passed to the Nuclei binary without explicit input validation in the Python code. While Nuclei is expected to handle its arguments safely, crafted inputs could potentially lead to unexpected behavior. The `SCAN_RESULTS_DIR` and `NUCLEI_BIN` paths are configurable via environment variables, which is standard but means their security depends on the environment's integrity. No 'eval', obfuscation, or hardcoded secrets were found. The tool is designed to run a powerful scanner, so inherent risks related to network traffic generation and resource usage should be acknowledged.
Updated: 2025-12-29GitHub→
0
0
Medium Cost

chictr-mcp-server

by PancrePal-xiaoyibao

Sec8

A Model Context Protocol (MCP) server for querying clinical trial information from the Chinese Clinical Trial Registry (ChiCTR).

Setup Requirements

  • ⚠️Requires Node.js and npm installed.
  • ⚠️Relies on external website (ChiCTR) availability and structure, subject to anti-bot measures; may require manual proxy configuration for stable operation.
  • ⚠️Requires an MCP client (e.g., Cherrystudio, @modelcontextprotocol/inspector) to interact with its functions, as it is an MCP server.
Verified SafeView Analysis
The server uses Playwright for web scraping, which involves making requests to an external, untrusted domain (ChiCTR website). While this is the intended functionality, it carries inherent risks associated with external network interactions and potential changes to the target site's content or structure. No 'eval' or obvious obfuscation found. Proxy support is provided for robust operation.
Updated: 2025-11-17GitHub→
0
0
Medium Cost
launchnotes icon

mcp

by launchnotes

Sec9

An MCP server for managing LaunchNotes projects, announcements, feedback, and analytics through the GraphQL API, designed for integration with clients like Claude Desktop.

Setup Requirements

  • ⚠️Requires `LAUNCHNOTES_API_TOKEN` environment variable.
  • ⚠️Requires Node.js version 18.0.0 or higher.
  • ⚠️A LaunchNotes 'Management token' is required for write operations; a 'Public token' may only grant read-only access.
Verified SafeView Analysis
The project demonstrates strong security practices: `LAUNCHNOTES_API_TOKEN` is correctly loaded from environment variables (no hardcoded secrets). All tool inputs are rigorously validated using Zod schemas, preventing common input-related vulnerabilities. GraphQL queries/mutations are used, and the client gracefully handles API errors (authentication, rate limits, timeouts). The server itself does not execute arbitrary code provided in custom HTML/CSS fields; it merely passes them to the LaunchNotes API, relying on the external API for final sanitization/rendering safety.
Updated: 2025-12-13GitHub→
0
0
Medium Cost
isakskogstad icon

Upphandlingsdata-MCP

by isakskogstad

Sec9

Provides access to Swedish public procurement data by integrating Upphandlingsmyndigheten APIs and TED (EU Tenders Electronic Daily).

Setup Requirements

  • ⚠️Requires Node.js version 18.0.0 or higher.
  • ⚠️Local setup requires specific configuration in Claude Desktop's claude_desktop_config.json, including the absolute path to the server's executable.
Verified SafeView Analysis
The server uses standard practices like input validation and limits request body size. It sanitizes upstream API URLs in error messages. No direct 'eval' or obvious malicious patterns found. No hardcoded secrets are evident, relying on public APIs. CORS is enabled for all origins, which is acceptable for a public data API.
Updated: 2025-12-02GitHub→
0
0
Medium Cost
HarryGremlins icon

altius-mcp-server

by HarryGremlins

Sec1

The server acts as an agent to interact with and manage specific Git repositories (revm, alloy, reth) by providing tools to list repositories, search code, read files, and list branches as part of the Model Context Protocol.

Setup Requirements

  • ⚠️Requires Git command-line tools to be installed and available in the execution environment.
  • ⚠️Requires environment variables `URL_REVM`, `URL_ALLOY`, and `URL_RETH` to be set to valid Git repository URLs for repository cloning.
  • ⚠️Requires internet access to clone and potentially pull updates for the specified repositories.
Review RequiredView Analysis
CRITICAL: The `search_code` tool is vulnerable to command injection. The `query` argument is directly interpolated into a `git grep` command without proper sanitization. An attacker can execute arbitrary shell commands. Potential command injection in `git clone` operations if environment variables `URL_REVM`, `URL_ALLOY`, `URL_RETH` are not trusted, as their values are directly used in shell commands. The `read_file` tool has basic path traversal prevention, but command injection for the `path` argument is still a risk if not fully sanitized.
Updated: 2026-01-19GitHub→
0
0
Low Cost
ShaySarussiElshten icon

weather-mcp-server

by ShaySarussiElshten

Sec9

Provides current weather conditions and 5-day forecasts for any city using the OpenWeatherMap API via a Model Context Protocol (MCP) server.

Setup Requirements

  • ⚠️Requires an OpenWeatherMap API Key (free tier available, but new keys may take up to 2 hours to activate).
  • ⚠️Requires Python 3.10+.
  • ⚠️Specific setup for integration with Claude Code (CLI tool or manual JSON configuration).
Verified SafeView Analysis
The server correctly retrieves the API key from environment variables, preventing hardcoding in the main application logic. It uses `httpx` for secure asynchronous HTTP requests. Input parameters are handled as API request parameters, minimizing injection risks. Error handling is present. A minor note is that the `test_server.py` script hardcodes an API key, which is acceptable for a local test but would be a risk if included in a public configuration or production deployment.
Updated: 2025-11-23GitHub→
0
0
Medium Cost
jeremyhahn icon

go-objstore

by jeremyhahn

Sec7

Provides a multi-protocol object storage client designed for integration with AI tools and LLMs, exposing storage operations as defined tools.

Setup Requirements

  • ⚠️Requires Docker and Docker Compose for easy setup and integration tests.
  • ⚠️Requires Go 1.21+ for local development and building the server components.
  • ⚠️For production use with cloud backends (S3, GCS, Azure), proper cloud credentials must be configured securely via environment variables.
Verified SafeView Analysis
The MCP server uses environment variables for sensitive cloud credentials (e.g., AWS, GCS, Azure), which is standard practice but requires secure handling during deployment. The integration tests use insecure configurations (e.g., MinIO/Azurite test credentials, self-signed QUIC certs) which are acceptable for testing but would be security risks in production. There are no immediate signs of 'eval' or similar dangerous patterns in the provided server code snippets. Potential risks exist if the server is exposed publicly without proper TLS certificates and access controls.
Updated: 2026-01-01GitHub→
0
0
Low Cost
dprince icon
Sec9

Managing and querying OpenStack Kubernetes custom resources (CRDs) via a Model Context Protocol (MCP) server.

Setup Requirements

  • ⚠️Go 1.22 or higher required for building
  • ⚠️Requires access to a Kubernetes cluster with OpenStack operators installed
  • ⚠️Requires valid Kubeconfig or in-cluster Kubernetes configuration
Verified SafeView Analysis
The server interacts with a Kubernetes cluster via the standard Kubernetes API client-go library, requiring valid kubeconfig or in-cluster configuration. Its security relies heavily on the Kubernetes cluster's RBAC setup and the permissions granted to the service account or user associated with the kubeconfig. There are no obvious hardcoded secrets, 'eval' functions, or direct network listeners exposed by the MCP server itself (it communicates over stdio).
Updated: 2025-11-26GitHub→
0
0
Low Cost
SethGame icon

mcp_flexsim

by SethGame

Sec3

Control FlexSim simulations using AI assistants via the Model Context Protocol (MCP) for digital twin integration and automated analysis.

Setup Requirements

  • ⚠️Requires Windows OS (FlexSim is Windows-only).
  • ⚠️Requires a local FlexSim installation and possibly an evaluation license.
  • ⚠️Python 3.12 is the default target version for FlexSimPy builds.
  • ⚠️Requires `uv` package manager for dependency management and script execution.
  • ⚠️Requires Microsoft Visual Studio with C++ development tools (MSBuild) to build the FlexSimPy C++ extension.
Review RequiredView Analysis
The `flexsim_evaluate` tool allows arbitrary FlexScript code execution via client requests, which can interact with the file system and FlexSim environment without apparent sanitization or access control. File operation tools (`flexsim_open_model`, `flexsim_save_model`, `flexsim_export_results`) accept arbitrary file paths, potentially leading to directory traversal or arbitrary file read/write vulnerabilities. The `flexsim_set_node_value` tool allows modification of arbitrary FlexSim model nodes, which could corrupt models or lead to unexpected behavior. This high level of client control over server-side actions, including script execution and file system access, presents significant security risks if the MCP client is untrusted or compromised.
Updated: 2025-11-30GitHub→
0
0
Medium Cost
ehlowr0ld icon
Sec9

The Scratch Notebook MCP server provides named scratchpads, typed cells, validation, namespaces, and semantic search to organize ideas for AI assistants in any MCP-enabled workspace.

Setup Requirements

  • ⚠️Requires Python 3.12+ (specified in pyproject.toml).
  • ⚠️Semantic search components (e.g., sentence-transformers and its PyTorch dependency) may download large models on first run, impacting initial startup time.
  • ⚠️The `--storage-dir` argument or `SCRATCH_NOTEBOOK_STORAGE_DIR` environment variable is crucial for specifying where persistent data (scratchpads, embeddings) is stored.
Verified SafeView Analysis
The server's source code demonstrates good security practices including explicit optional bearer-token authentication for HTTP/SSE transports (defaulting to local-only binding without auth), safe YAML parsing (`yaml.safe_load`), and validation of configuration inputs. It does not appear to use dangerous dynamic code execution functions like `eval()` on arbitrary user input. The primary security consideration for operators is the `uvx` installation method which fetches code from a Git repository, requiring trust in the specified source to avoid supply chain risks.
Updated: 2025-11-21GitHub→
0
0
Medium Cost
oluwaeinstein007 icon

MegaMind

by oluwaeinstein007

Sec1

A content ingestion and processing system for AI applications, capable of crawling web pages, parsing documents, chunking text, generating LLM embeddings, and storing data for semantic search.

Setup Requirements

  • ⚠️Requires paid LLM API Key (OpenAI or Google Gemini) for embeddings.
  • ⚠️Requires a Qdrant vector database instance (self-hosted or cloud-based) if `QDRANT_ENABLED=true`.
  • ⚠️Native C++ build tools required for `better-sqlite3` (e.g., Python, C++ compiler) for persistent storage; otherwise, it falls back to in-memory storage.
  • ⚠️Requires Playwright browsers to be installed (`npx playwright install`) for web crawling.
Review RequiredView Analysis
CRITICAL: The MCP server exposes tools like `INGEST_URL_TOOL` and `INGEST_FILE_TOOL` that accept arbitrary URLs for web crawling and arbitrary file paths for ingestion directly from the MCP client. This creates severe security vulnerabilities: 1. Server-Side Request Forgery (SSRF): An attacker could use `INGEST_URL_TOOL` to force the server to make requests to internal network resources, potentially disclosing sensitive information or exploiting internal services. 2. Local File Inclusion/Disclosure: An attacker could use `INGEST_FILE_TOOL` with paths like `/etc/passwd` or `../../.env` to read and ingest sensitive files from the server's filesystem. These tools lack explicit input validation or sanitization within the provided code, making them highly dangerous if exposed to untrusted input. The `transportType: 'stdio'` might mitigate direct network exposure, but a compromised MCP client or malicious input via the stdio channel still poses these risks.
Updated: 2025-12-11GitHub→
PreviousPage 453 of 760Next