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)

30
1
High Cost
jordnlvr icon
Sec4

A self-learning Model Context Protocol (MCP) server providing expert-level Mendix development assistance, including knowledge base queries, project and theme analysis, code generation patterns, and troubleshooting, with support for semantic search and auto-harvesting.

Setup Requirements

  • ⚠️Requires OpenAI or Azure OpenAI API Key (Paid) for high-quality vector embeddings, falls back to local TF-IDF if not configured.
  • ⚠️Requires Pinecone API Key (Free tier available, but usage costs apply) for semantic search.
  • ⚠️Requires Supabase account (Free tier available) for persistent cloud knowledge storage.
  • ⚠️Project analysis (`/analyze`, `/analyze-theme`) functionality is disabled in cloud deployments due to requiring local filesystem access to `.mpr` files or project directories.
  • ⚠️The `SyncReminder` module relies on `git` being installed and configured locally for repository synchronization tasks.
Review RequiredView Analysis
The project includes an obfuscated, hardcoded Pinecone API key in `src/vector/VectorStore.js` as a fallback if environment variables are not set. This is a critical security vulnerability as it could expose access to a shared Pinecone index. While `flyio-secrets.sh` indicates proper environment variable usage for deployment, a local run without explicit configuration would use the hardcoded key. The `SyncReminder` module uses `child_process.execSync` for git operations, which carries inherent risks if not carefully managed, although its current use for self-repo management appears controlled. Public REST and SSE endpoints lack user-level authentication, relying on rate-limiting for abuse prevention, which is noted as a future improvement in the `ARCHITECTURE.md`.
Updated: 2026-01-19GitHub→
30
1
Low Cost
woonstadrotterdam icon

mcp-monumenten

by woonstadrotterdam

Sec9

Enables AI assistants to check monumental status of Dutch addresses.

Setup Requirements

  • ⚠️Requires Python 3.10 or higher.
Verified SafeView Analysis
The server uses f-strings to build SPARQL queries for the external Kadaster BAG LV endpoint. User inputs for street, city, postal code, house letter, and house suffix are properly enclosed in double quotes within the SPARQL query, which effectively mitigates classical SPARQL injection risks by treating inputs as literal strings. No 'eval' or other direct command execution patterns are found. The server relies on external Kadaster and RCE APIs for data, whose security is external to this codebase. No hardcoded secrets are evident, and `dotenv` is used for environment variable loading, suggesting sensitive data should be managed externally.
Updated: 2025-11-30GitHub→
30
1
Low Cost
Sunako26 icon
Sec7

Manages Docker, Podman, and Docker Swarm containers via a centralized Multi-Agent Communication Protocol (MCP) server for Agentic AI.

Setup Requirements

  • ⚠️Requires Docker or Podman daemon to be installed and running on the host system.
  • ⚠️Python 3.10 or higher is required.
  • ⚠️If using Podman on Windows/WSL2, 'podman machine' must be running and accessible.
Verified SafeView Analysis
The server allows external agents to execute powerful container management commands, including running arbitrary images and executing commands inside containers. While the code itself does not show obvious internal vulnerabilities like 'eval' or hardcoded secrets, the inherent capabilities grant high privileges. If the MCP server or an agent connected to it is compromised, it could lead to significant control over the host's container environment. Deployment security (network isolation, trusted agents) is critical.
Updated: 2026-01-19GitHub→
30
1
Medium Cost
David200308 icon
Sec8

Provides cryptocurrency price checking across multiple exchanges and sends email alerts.

Setup Requirements

  • ⚠️Requires Node.js >=18.0.0 and pnpm >=10.0.0.
  • ⚠️Requires a Resend API Key for email functionality (RESEND_API_KEY).
  • ⚠️Optional API keys (COINMARKETCAP_API_KEY, MORALIS_API_KEY, ETHERSCAN_API_KEY) can enhance token address resolution but are not strictly required for core functionality.
Verified SafeView Analysis
API keys for external services (Resend, CoinMarketCap, Moralis, Etherscan) are correctly loaded from environment variables or a configuration file, not hardcoded. The server can run as an HTTP service using Fastify, which is a robust framework. Outbound network requests are made to various crypto data providers and blockchain RPCs. A potential minor risk is the `convertMarkdownToHtml` function in `EmailService`: if the `content` or `additionalInfo` parameters for email templates were to receive unsanitized, malicious markdown/HTML directly from an untrusted source (e.g., a vulnerable LLM output), it could potentially lead to XSS vulnerabilities in the recipient's email client. However, this is a common risk with HTML templating and requires careful input handling by the calling agent.
Updated: 2025-12-05GitHub→
30
1
High Cost
wolfeidau icon

mcp-evals

by wolfeidau

Sec7

A Go library and CLI for evaluating Model Context Protocol (MCP) servers using Claude, employing an agentic evaluation loop and detailed grading across multiple dimensions.

Setup Requirements

  • ⚠️Requires an Anthropic API Key (Paid Service) for LLM interactions.
  • ⚠️Requires a Go development environment (Go 1.25+) to build from source, or relies on `install.sh` for pre-built binaries.
  • ⚠️Requires a functional target MCP server (either a local executable or a network service) that the tool will connect to and evaluate. This target server's dependencies (e.g., Node.js for `npx @modelcontextprotocol/server-filesystem`) must also be met.
Verified SafeView Analysis
The tool is designed to execute external commands specified in its configuration files (e.g., `mcp_server.command` and `mcp_server.args`). While this is necessary functionality for evaluating MCP servers, it means that if the configuration file is untrusted, the `mcp-evals` tool could execute arbitrary malicious commands on the host system. The `install.sh` script handles binary downloads and checksum verification appropriately. No hardcoded secrets, obfuscation, or direct shell `eval` calls are found in the Go source code.
Updated: 2026-01-19GitHub→
30
1
Medium Cost
danielrosehill icon

MCP-Server-Ideas

by danielrosehill

Sec9

Planning and documentation for Model Context Protocol (MCP) servers to extend AI capabilities by integrating with external APIs such as Green Invoice for invoicing and Homebox for inventory management.

Setup Requirements

  • ⚠️Requires API keys for Green Invoice API (Paid/Account needed)
  • ⚠️Requires API keys for Homebox API (Paid/Account needed)
Verified SafeView Analysis
The repository consists primarily of planning documents and an OpenAPI schema, not executable code. Therefore, there are no direct runtime security risks in the provided source. Implementation notes within the planning documents mention the need for robust data privacy, security, and API key management, indicating an awareness of security for future development.
Updated: 2025-12-02GitHub→
30
1
Medium Cost
ScientiaCapital icon

unsloth-mcp-server

by ScientiaCapital

Sec8

An MCP server enabling Claude Code users to orchestrate LLM fine-tuning workflows, manage GPU resources, and generate training data, leveraging Unsloth's performance optimizations.

Setup Requirements

  • ⚠️Requires Python 3.10-3.12 (not 3.13).
Verified SafeView Analysis
The server implements comprehensive security measures including input validation, path traversal prevention, and Python script sanitization against dangerous patterns (e.g., 'os.system', 'eval', 'subprocess', file writes). Resource limits and rate limiting are in place. Secrets are handled via environment variables. Detected dangerous patterns in Python scripts are currently logged as warnings, not errors, which could be tightened for maximum security in a strict production environment.
Updated: 2026-01-18GitHub→
30
1
Medium Cost
kuc-arc-f icon

mcp_server_2ex

by kuc-arc-f

Sec2

This server acts as a Next.js application providing a Remote Procedure Call (RPC) interface for Retrieval Augmented Generation (RAG) search using PostgreSQL with pgvector for vector storage and Google Generative AI for language model inference.

Setup Requirements

  • ⚠️Requires PostgreSQL with pgvector extension running.
  • ⚠️Requires Google API Key (Paid service).
  • ⚠️Weak default database credentials (`PG_PASSWORD=admin`) in environment variable examples.
  • ⚠️The /api/mcp endpoint lacks any authorization, making it publicly accessible.
Review RequiredView Analysis
The primary API endpoint (`/api/mcp`) lacks an authorization mechanism, making it openly accessible to anyone on the network. The provided `.env` example suggests weak default database credentials (`PG_PASSWORD=admin`), which pose a significant security risk if used in production.
Updated: 2026-01-10GitHub→
30
1
Low Cost
eduoncode icon

mcp_server

by eduoncode

Sec8

This server implements the Model Context Protocol (MCP) to expose developer-defined tools as a HTTP API, specifically demonstrating a task creation tool.

Setup Requirements

  • ⚠️Requires environment variables (e.g., PORT) to be set via .env file or system environment.
  • ⚠️Requires a TypeScript build step (`tsc`) before running in production, or `tsx` for development.
Verified SafeView Analysis
The server uses the `@modelcontextprotocol/sdk` to implement an MCP server over HTTP. It utilizes `dotenv` for environment variable loading (e.g., PORT), which is good practice for configuration. The exposed `create_task` tool, as defined, is simple and primarily logs input and returns a formatted string, posing no direct security risks like command injection or unauthorized file access. No `eval` or similar dangerous patterns are directly visible in the provided source code. Security largely depends on the robustness of the underlying `@modelcontextprotocol/sdk` and proper environment configuration.
Updated: 2025-12-02GitHub→
30
1
Medium Cost
23prime icon
Sec9

Accesses and analyzes OEIS (Online Encyclopedia of Integer Sequences) data via the Model Context Protocol.

Setup Requirements

  • ⚠️Docker is recommended for easy deployment.
  • ⚠️Development requires Taskfile and mise.
Verified SafeView Analysis
The server binds to "0.0.0.0" by default, which is common for containerized applications but requires appropriate network configuration (e.g., firewalling) in production. It makes external HTTP GET requests to "https://oeis.org" to fetch data; reliance on external services always carries a minor inherent risk if the external service is compromised or behaves maliciously, though the data processing within this server appears robust, using typed structures and JSON serialization.
Updated: 2026-01-16GitHub→
30
5
High Cost
docentovich icon

chrometools-mcp

by docentovich

Sec3

MCP server for browser automation, web scraping, and UI testing using Puppeteer, enhanced with AI-powered element finding, page analysis, visual UI recording, and Figma design-to-code validation.

Setup Requirements

  • ⚠️Requires Node.js >= 18.0.0.
  • ⚠️For GUI mode in Windows Subsystem for Linux (WSL), it requires VcXsrv Windows X Server on Windows with 'Disable access control' enabled, and the `DISPLAY` environment variable configured to the Windows host IP (e.g., `DISPLAY=172.25.96.1:0`).
  • ⚠️Figma-related tools require a Figma Personal Access Token, preferably set via the `FIGMA_TOKEN` environment variable.
Review RequiredView Analysis
The server includes an `executeScript` tool that uses `eval()` to execute arbitrary JavaScript code provided by the user (or AI agent) directly in the browser's context. Similarly, `new Function()` is used in `evaluateCustomCondition` within the recorder. This capability, while central to browser automation, represents a significant security risk if the input is not from a fully trusted source, potentially allowing for cross-site scripting (XSS), data exfiltration, or other malicious actions if the MCP client or its input is compromised. There are no obvious hardcoded secrets beyond configuration keywords, and API tokens are managed via environment variables.
Updated: 2026-01-05GitHub→
30
1
High Cost
janisz icon

sejm-mcp

by janisz

Sec9

Serves as a Model Context Protocol (MCP) server providing AI assistants with access to real-time Polish parliamentary data (Sejm API) and legal documents (European Legislation Identifier - ELI API).

Setup Requirements

  • ⚠️Requires Go 1.21+ for building from source (development setup).
  • ⚠️Relies on external Polish government APIs (api.sejm.gov.pl) for all data, requiring active internet connectivity.
  • ⚠️Docker is the recommended deployment method, adding a dependency on Docker runtime.
Verified SafeView Analysis
The codebase is clean, uses standard and well-vetted Go libraries, and performs robust input validation and error handling for external API calls. No instances of 'eval' or similar dynamic code execution functions were found. No hardcoded sensitive secrets are present. The external API endpoints it consumes are public. PDF parsing is handled by the `gen2brain/go-fitz` library, a common choice for such tasks.
Updated: 2026-01-12GitHub→
PreviousPage 229 of 760Next