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
shiveshnavin icon

pipelane-server

by shiveshnavin

Sec1

A server for building, scheduling, and managing custom data pipelines, integrating with AI agents via the Model Context Protocol (MCP) and providing a GraphQL API for interaction.

Setup Requirements

  • ⚠️Requires `sqlite3` npm package for SQLite persistence, or connection details for MySQL.
  • ⚠️The client-side UI needs to be built using `cd client && npm run build` before the server can be fully started.
  • ⚠️Integration with AI models (e.g., Google Gemini) requires setting the `GEMINI_API_KEY` environment variable.
  • ⚠️For HTTPS, `PIPELANE_HTTPS_KEY_PATH` and `PIPELANE_HTTPS_CERT_PATH` environment variables pointing to valid SSL certificate files are required.
Review RequiredView Analysis
Critical security risks identified: 1. Direct use of `eval(js)` in `EvaluateJsTask` and related functions (`evalPlaceHolder.evaluatePlaceholdersInString`, `evalPlaceHolder.evalInScope`) allows arbitrary JavaScript code execution from user-defined pipeline inputs or task configurations. This is a severe Remote Code Execution (RCE) vulnerability. 2. `ShellTask` and `LoopShellTask` allow execution of arbitrary shell commands. The `allowedCommands` filter, as configured in `server/pipe-tasks/index.ts`, can include `"*"`, effectively bypassing all restrictions and enabling RCE via shell injection. These vulnerabilities make the system unsafe for untrusted users or without robust sandboxing.
Updated: 2026-01-19GitHub→
0
0
High Cost
zenyrae123 icon
Sec9

Automated, multi-dimensional e-commerce data analysis and business intelligence reporting.

Setup Requirements

  • ⚠️Requires Python 3.8+ with specific data science packages (pandas, numpy, scipy, matplotlib, seaborn).
  • ⚠️Designed to run within a Claude Code environment, leveraging its sub-agents, skills, and slash commands (`/do-more`, `/do-all`).
  • ⚠️Relies on local data files located in a `data_storage/` directory for analysis.
Verified SafeView Analysis
The project extensively uses standard data science libraries (pandas, numpy, scikit-learn, matplotlib, seaborn). There are no direct malicious patterns, `eval` of arbitrary user input, or hardcoded sensitive credentials observed in the provided code snippets. The documentation explicitly mentions security practices like `validate-analysis` hooks to check for dangerous code patterns (e.g., `os.system`, `eval`, `exec`, `subprocess`) and `allowed-tools` to limit skill permissions, indicating a focus on security within the Claude Code environment.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
LanceVCS icon

gemini-mcp

by LanceVCS

Sec9

Provides multi-turn conversational support for the Gemini CLI, allowing a client application (like Claude Code) to maintain context across multiple interactions with Gemini.

Setup Requirements

  • ⚠️Requires Gemini CLI to be installed and configured (including authentication and potential API key setup, which usually involves costs).
  • ⚠️Requires Node.js 14.0 or higher.
  • ⚠️Conversations auto-expire after 2 hours or when the maximum of 25 concurrent conversations is reached.
Verified SafeView Analysis
The server spawns the 'gemini' CLI process using `child_process.spawn`. Critical security mitigations include: 1) Explicit whitelisting of 'SAFE_TOOLS' which notably *excludes* dangerous commands like `run_shell_command`, `browser_run_code`, `browser_evaluate`, etc. 2) The `gemini` CLI is run with the `--sandbox` flag, which restricts file system access. 3) User prompts are passed as arguments to the `gemini` CLI, preventing direct shell injection by using an array of arguments rather than a single concatenated string. No 'eval' or direct arbitrary code execution is present within the MCP server's logic itself. The primary remaining risk lies with potential vulnerabilities within the Gemini CLI itself, particularly how it processes input when run with the specified arguments and sandbox restrictions.
Updated: 2025-12-17GitHub→
0
0
Medium Cost
sakkkysaka icon

mcp-openfoodtox

by sakkkysaka

Sec9

Access, search, and analyze EFSA's OpenFoodTox chemical hazards dataset to explore food safety information, toxicity endpoints, risk assessments, and substance safety profiles.

Setup Requirements

  • ⚠️Requires Python 3.12 or later.
  • ⚠️Requires 'uv' for Python package management.
  • ⚠️The local SQLite database (`openfoodtox.db`) must be populated, which depends on the availability of the large source Excel file (`OpenFoodToxTX22809_2023.xlsx`) for initial setup.
Verified SafeView Analysis
The server uses `sqlite3` and `pandas.read_sql_query` for database interactions. All queries observed utilize parameterized inputs (`params=...`), which effectively prevents SQL injection vulnerabilities. No `eval` or `exec` functions were found, nor were any hardcoded secrets or malicious patterns detected. The `FastMCP` server is configured to run with `transport='stdio'`, implying local communication typically used for LLM agent integration, which limits its direct network attack surface. However, `httpx` is a dependency, and the README mentions 'Active internet connection for updates and dataset access', suggesting external network calls might occur as part of a broader system or initial setup not directly within the provided server runtime logic.
Updated: 2026-01-19GitHub→
0
0
High Cost
mesolimbo icon

gemini-mcp

by mesolimbo

Sec9

A Model Context Protocol (MCP) server providing local access to Google's Gemini 3 Pro Preview API for integration with tools like Claude Code.

Setup Requirements

  • ⚠️Requires a Google Gemini API Key (obtain from aistudio.google.com/apikey)
  • ⚠️Requires Node.js version >= 22.21.0
  • ⚠️Requires creation and editing of `config.json` from `config.json.example` to provide the API key.
Verified SafeView Analysis
The server uses `config.json` for the Gemini API key, which is explicitly `gitignored` as per best practices. It does not contain `eval` or obvious obfuscation. Network interaction is limited to calls to the Google Gemini API. It runs locally via stdio, not exposing external network ports, which is generally secure. Configuration validation for the API key is present at startup.
Updated: 2026-01-07GitHub→
0
0
Medium Cost
apkmethedisalpha icon

note-com-mcp

by apkmethedisalpha

Sec7

This server acts as a Model Context Protocol (MCP) gateway to the note.com API, enabling AI agents and automation tools like n8n, Cursor, and ChatGPT to search, analyze, and manage notes on the note.com platform.

Setup Requirements

  • ⚠️Requires note.com account credentials (email, password) or session cookies/tokens (`NOTE_SESSION_V5`, `NOTE_XSRF_TOKEN`) to be set in a `.env` file for authenticated operations. A script is provided to help capture these credentials.
  • ⚠️For remote access and robust integration (e.g., with n8n on a VPS), additional setup is required for secure exposure, such as configuring Cloudflare Tunnel or Nginx with HTTPS and basic authentication. The HTTP server defaults to `localhost` binding.
  • ⚠️The server is a Node.js application, requiring a Node.js runtime (v16+) to be installed.
Verified SafeView Analysis
The server primarily relies on environment variables (.env file) for sensitive data like note.com session cookies, XSRF tokens, email, and password. While this prevents hardcoding in the source, it shifts the responsibility of securing the .env file to the user. The HTTP server's default CORS setting is '*' (allow all origins), which could be a risk if exposed directly to the internet without a protective proxy (like Nginx or Cloudflare Tunnel) with stricter access controls. Debug logs can output sensitive API responses and request bodies if `DEBUG` is enabled. The `degenerator` dependency, used for PAC file resolution, involves dynamic code generation, but within a standard, isolated context not directly exposed to arbitrary user input.
Updated: 2025-12-15GitHub→
0
0
Medium Cost
nf-osi icon

mcp-server

by nf-osi

Sec5

Provides tools for Synapse data curation workflows, including portal metadata generation and project review/classification, enabling AI assistants to interact with Synapse resources.

Setup Requirements

  • ⚠️Requires `SYNAPSE_AUTH_TOKEN` environment variable, a personal access token for Synapse (get from https://www.synapse.org/ -> Account Settings -> Personal Access Tokens).
  • ⚠️Requires Python 3.10+.
  • ⚠️Recommended installation and running involves `uvx`, a specific Python dependency management tool.
Review RequiredView Analysis
The server exposes powerful operations directly to the LLM that involve user-provided strings without robust internal validation: 1. SQL Injection Risk: The `synapse_query` tool passes raw `query` strings directly to `synapseclient.tableQuery`. While Synapse may offer some backend protection, an LLM receiving malicious input could craft queries to extract sensitive data or perform unintended operations within the bounds of the `SYNAPSE_AUTH_TOKEN`'s permissions. 2. Arbitrary File Read Risk: The `validate_metadata` tool accepts a `schema_file` path from user input. Without proper sanitization or sandboxing by the MCP client, this could allow the LLM to read arbitrary files from the server's filesystem. 3. Server-Side Request Forgery (SSRF) Risk: Tools like `fetch_schema`, `get_data_classes`, and `get_data_sharing_plan` construct URLs based on user input (e.g., `schema_url`, `templates_url`, `study_id` for DSP API). An LLM could potentially craft these inputs to probe or interact with internal network services if the server has such access. 4. Broad Permissions: The required `SYNAPSE_AUTH_TOKEN` is a personal access token, which can grant extensive permissions to Synapse resources, amplifying the impact of any successful exploitation of the above risks.
Updated: 2025-11-24GitHub→
0
0
Low Cost

mcp-wizard

by joewxboy

Sec3

A utility to assist with Minecraft Coder Pack (MCP) setup or development tasks, potentially simplifying the modding process.

Review RequiredView Analysis
A detailed security audit is impossible as the 'README Truncated: 404: Not Found' error indicates no code or sufficient context is available for analysis. Running any software, especially a 'wizard' utility that might interact with system files or environments, without source code review or documentation, presents a significant and unquantifiable security risk.
Updated: 2025-11-18GitHub→
0
0
Low Cost
IhorHolovatsky icon

iholovatsky-mcps

by IhorHolovatsky

Sec7

This repository provides Model Context Protocol (MCP) servers to integrate various external services (financial, business) into personal AI workflows, acting as data providers for AI clients.

Setup Requirements

  • ⚠️Requires .NET 10 SDK
Verified SafeView Analysis
The server components communicate via STDIO, limiting direct network exposure to local processes, which is generally safer. However, the actual C# source code for the integrations (e.g., with financial institutions like IBKR and BNP Paribas) is not provided, making it impossible to audit for hardcoded secrets, improper data handling, or other vulnerabilities within the application logic itself. The inherent nature of these integrations suggests handling sensitive data.
Updated: 2025-11-24GitHub→
0
0
Medium Cost
ajinux icon
Sec8

Integrates Pi-hole DNS query data with AI-powered OSINT for home network monitoring and suspicious domain investigation.

Setup Requirements

  • ⚠️Requires a running Pi-hole instance.
  • ⚠️Requires the Pi-hole API password.
  • ⚠️Needs an MCP client to interact with, as it's an MCP server, not a standalone UI.
Verified SafeView Analysis
The server correctly handles sensitive information by requiring the Pi-hole API password via environment variables, .env file, or command-line flags. It uses standard Go libraries and well-known third-party packages for DNS querying and WHOIS lookups. There are no instances of 'eval' or other dynamic code execution risks. The installation script downloads a binary from GitHub, which carries standard supply chain risks if the repository is compromised, but the script itself is straightforward. Network connections are made to Pi-hole (internal) and public DNS/WHOIS servers (external) as part of its intended functionality.
Updated: 2025-12-14GitHub→
0
0
Low Cost
fwornle icon
Sec8

A real-time constraint enforcement and monitoring system designed to integrate with AI coding agents (like Claude Code) to prevent policy violations and ensure code quality and security before execution. It provides a dashboard for visualization and management of constraints and violations.

Setup Requirements

  • ⚠️Claude Code Hook Configuration: Requires manual or semi-automated modification of Claude Code's local settings file (.claude/settings.local.json) and a restart of Claude Code to activate real-time constraint enforcement hooks (UserPromptSubmit, PreToolUse).
  • ⚠️LLM API Key Requirement: Semantic validation, a core advanced feature, requires an API key for a supported LLM provider (e.g., GROK_API_KEY for Groq, ANTHROPIC_API_KEY for Anthropic, GOOGLE_API_KEY for Gemini), which may incur costs and require external accounts.
  • ⚠️External Database Dependencies: Certain features (vector search for violations, advanced analytics) rely on external databases like Qdrant (vector DB) and Redis (cache), often deployed via Docker. The setup script warns if QDRANT_HOST is not set and suggests a docker run command.
Verified SafeView Analysis
The server's primary function is to enhance security by enforcing coding constraints (e.g., detecting hardcoded secrets, 'eval()' usage). It uses 'execSync' internally for loading configurations and health checks, which, while controlled by internal paths, could be a vector if external inputs were to influence the commands. The dashboard API is exposed with CORS for all origins, which is appropriate for local development but a significant risk if publicly exposed. Logging is carefully configured to avoid stdout when running as an MCP server or hook, preventing RPC stream corruption.
Updated: 2026-01-10GitHub→
0
0
Low Cost
sebastienblanc icon

mcp-api-server

by sebastienblanc

Sec9

Provides an authenticated API and internal tools for managing and assessing database regions within a platform.

Setup Requirements

  • ⚠️Requires Java Development Kit (JDK) 11+ (or higher, depending on Quarkus version).
  • ⚠️Requires Maven or Gradle to build and run the Quarkus application.
  • ⚠️Requires an authentication provider (e.g., OIDC, JWT) configured for Quarkus to satisfy '@Authenticated' and '@RolesAllowed' annotations.
Verified SafeView Analysis
The code correctly implements authentication and role-based authorization for its API endpoints and tools using Quarkus security features. The current implementations are simple string echoes and logging, showing no direct interaction with databases or system commands that could introduce injection vulnerabilities within this snippet. No 'eval', obfuscation, or hardcoded secrets are present. The actual security of region management depends on external services not detailed in this snippet.
Updated: 2025-12-15GitHub→
PreviousPage 309 of 760Next