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
Hao17 icon
Sec7

Provides AI systems programmatic access to Ghidra's reverse engineering capabilities via an HTTP JSON API.

Setup Requirements

  • ⚠️Requires Ghidra to be running; the `ghidra_mcp_server.py` script must be executed within Ghidrathon.
  • ⚠️Requires `pip install mcp uvicorn httpx` for the MCP SSE proxy functionality.
  • ⚠️The Python interpreter path for the MCP SSE proxy (`scripts/mcp_sse_proxy.py`) might need to be explicitly set via the `GHIDRA_MCP_PYTHON` environment variable if Ghidrathon's embedded Python is not accessible to subprocesses.
Verified SafeView Analysis
The server binds to localhost (127.0.0.1) by default, limiting network exposure. The use of `eval` in `api/demo.py` is contained within hardcoded, known strings, making it non-exploitable as implemented. `subprocess.Popen` is used to run an SSE proxy, which is a legitimate architectural pattern. The core functionality involves passing HTTP query parameters directly to Ghidra's Java APIs via Ghidrathon; the security relies heavily on the robustness of Ghidra's APIs against malicious input, which could potentially lead to crashes or unexpected behavior within the Ghidra environment. However, direct remote code execution on the underlying system is not an apparent risk due to these design choices.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
t-neumann icon

TxGemma-MCP

by t-neumann

Sec1

Deploying the TxGemma model as a local MCP server for AI/ML inference.

Setup Requirements

  • ⚠️Gemma model files must be downloaded and configured for the server.
  • ⚠️Requires sufficient compute resources (CPU/GPU, RAM) for running the Gemma model locally.
Review RequiredView Analysis
No source code (beyond README.md) was provided for analysis. Therefore, a security audit cannot be performed for patterns like 'eval', obfuscation, network risks, hardcoded secrets, or malicious patterns. The server's safety cannot be verified.
Updated: 2026-01-19GitHub→
0
0
Low Cost
FaisalAhmedBijoy icon

AI-Agent-MCP-Server

by FaisalAhmedBijoy

Sec3

Serves as a backend for AI agents, likely in the context of Minecraft (MCP).

Review RequiredView Analysis
Only the README.md file was provided as source code, preventing a comprehensive security audit for 'eval', obfuscation, network risks, hardcoded secrets, or malicious patterns. The safety of running this project cannot be determined without access to its actual implementation.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
HeshamWM icon

idfm-mcp-server

by HeshamWM

Sec9

Provides real-time Paris public transport data (Île-de-France Mobilités) to AI assistants via an MCP server.

Setup Requirements

  • ⚠️Requires Docker Desktop with MCP Toolkit enabled.
  • ⚠️Requires a PRIM API key from prim.iledefrance-mobilites.fr (registration likely required, possibly paid tiers).
  • ⚠️Uses Python 3.11.
Verified SafeView Analysis
The server correctly handles API keys by fetching them from environment variables or Docker-mounted secret files, preventing hardcoding. It uses `httpx` for HTTP requests, which is standard. No `eval` or other dangerous code execution patterns are observed. The container is designed to run as a non-root user. Overall, it follows good security practices for an API wrapper.
Updated: 2025-11-19GitHub→
0
0
Low Cost
serser152 icon

planner_mcp

by serser152

Sec9

A Micro-Capability Protocol (MCP) server designed to serve as a planning tool for an intelligent agent, providing task management and time/date functionalities.

Setup Requirements

  • ⚠️Requires Docker for the recommended installation and execution method.
  • ⚠️The 'add_task' function, although defined in 'mcp_server.py' and mentioned in the README's function list, is not decorated with '@mcp.tool()' and therefore not exposed as an accessible tool for agents via the MCP server.
  • ⚠️The 'get_current_time' function hardcodes a '+3 hours' timedelta, potentially leading to unexpected time values if the agent/user expects UTC or a different local timezone.
Verified SafeView Analysis
The server uses SQLAlchemy for all database interactions, which properly parameterizes queries and mitigates SQL injection risks. No 'eval' or direct shell command execution with user input was identified. No hardcoded sensitive credentials were found. The server binds to '0.0.0.0' for network access, which is standard for container deployments but requires appropriate host firewall configuration for external exposure.
Updated: 2025-11-26GitHub→
0
0
Medium Cost
Clauskraft icon
Sec9

Serves as a universal API gateway for Claude Desktop, providing access to various external services like social media, search engines, and LLMs through the Model Context Protocol (MCP).

Setup Requirements

  • ⚠️Requires API keys for external services (Twitter, Google, OpenAI/LLM, Brave), which may incur costs or rate limits.
  • ⚠️Requires Node.js/npm or Docker runtime environment.
  • ⚠️Designed for integration with Claude Desktop.
Verified SafeView Analysis
The provided `src/index.ts` is a stub that defines tools but does not implement their actual functionality to call external APIs (Twitter, Google, LLM, Brave). It currently only logs tool calls and returns a dummy string. No 'eval' or direct network requests are present in this specific file. Environment variables are used for API keys, which is good practice. If the actual API integration logic were present, a deeper security analysis of those integrations (e.g., handling credentials, input sanitization for external calls) would be necessary.
Updated: 2025-11-18GitHub→
0
0
Medium Cost
vibeatlas icon

ship-mcp-server

by vibeatlas

Sec9

Provides AI coding reliability scores and feedback mechanisms for AI coding assistants like Claude Desktop and Cursor by integrating with the SHIP Protocol API.

Setup Requirements

  • ⚠️Requires Node.js (version 18 or higher) and npm/npx to be installed on the host machine.
  • ⚠️Requires manual configuration in the settings of AI coding environments (e.g., Claude Desktop or Cursor).
  • ⚠️Optional `SHIP_API_KEY` can be provided for personalized predictions, which may involve usage costs or registration with the external SHIP Protocol service.
Verified SafeView Analysis
The server's source code is clear and does not exhibit direct use of dangerous functions like 'eval', code obfuscation, or hardcoded secrets. It utilizes environment variables for sensitive data like API keys. The primary security consideration lies in its reliance on an external SHIP Protocol API (ship.vibeatlas.dev). While the client code itself appears robust, the security and data handling practices of the external API are critical and should be trusted by the user.
Updated: 2025-12-06GitHub→
0
0
Medium Cost
multiluca2020 icon

visum-thinker-mcp-server

by multiluca2020

Sec1

AI-driven sequential thinking, large PDF processing, and PTV Visum transportation modeling analysis and automation.

Setup Requirements

  • ⚠️Requires Windows operating system due to heavy reliance on PowerShell and PTV Visum COM API.
  • ⚠️Requires PTV Visum (defaulting to 2025 64-bit) to be installed on the system, including its Python API (VisumPy).
  • ⚠️Many hardcoded file paths (e.g., 'H:\Program Files\PTV Vision\PTV Visum 2025\Exe\Python\python.exe', 'H:\go\italferr2025\Campoleone\...') necessitate adaptation to the specific deployment environment.
  • ⚠️Processing large PDF files and complex Visum projects can be memory and disk intensive.
Review RequiredView Analysis
CRITICAL VULNERABILITY: The server executes dynamically generated or user-provided Python code directly using `exec()` within a privileged Python environment that has access to the PTV Visum COM object. This allows for arbitrary code execution, which is extremely dangerous. Any malicious input to tools like `visum_custom_analysis`, `project_execute_analysis`, or `project_execute` could compromise the entire system if not properly sandboxed or strictly sanitized. Given the COM object access, this could lead to system-level compromise on the host machine.
Updated: 2026-01-18GitHub→
0
0
Low Cost
111031075jack icon

mcp-server-9090

by 111031075jack

Sec9

Provides a shopping cart and inventory management service designed to be exposed as tools for an AI orchestrator or LLM.

Setup Requirements

  • ⚠️Requires Java Development Kit (JDK) to compile and run.
  • ⚠️Intended for integration with a Spring AI orchestrator/LLM to utilize its '@Tool' annotated methods.
Verified SafeView Analysis
The provided source code is clean, without 'eval', obfuscation, or hardcoded secrets. It implements simple in-memory data structures for product catalog and cart. There are no direct database interactions shown. As a Spring Boot application, standard web security practices (authentication, authorization, advanced input validation) would be necessary for a production environment, but are not within the scope of the provided snippets.
Updated: 2025-12-13GitHub→
0
0
Medium Cost
ywatanabe1989 icon

sci-writer

by ywatanabe1989

Sec3

Manages and compiles scientific LaTeX manuscripts, including content creation, document analysis, versioning, and background job processing for AI assistants.

Setup Requirements

  • ⚠️Requires the `fastmcp` Python package.
  • ⚠️Requires a full LaTeX distribution (e.g., `texlive-full`, ~5GB) OR a container runtime (`Apptainer`/`Singularity`).
  • ⚠️Requires `bibtexparser`, `rich`, `click`, `pydantic` Python packages.
Review RequiredView Analysis
The server enables LaTeX's `\write18` (shell escape) feature during compilation (`pdflatex -shell-escape` in compilation scripts). This means that any content submitted to section/figure/table handlers by an AI agent (or other client) can potentially execute arbitrary shell commands if it contains `\write18{malicious_command}`. While designed for trusted AI agents creating LaTeX content, this is a severe code execution vulnerability if deployed in an untrusted environment or if the AI agent itself is compromised. Pydantic validation for input content ensures type safety but does not filter for potentially malicious LaTeX commands. Direct `eval` calls in shell scripts are used with internally constructed commands, which reduces direct command injection risk through function arguments, but the `\write18` vector remains. The system also pulls external containers and packages, which introduces supply chain risks.
Updated: 2026-01-18GitHub→
0
0
High Cost
OI-OS icon
Sec9

A vivid backend for AI agents, providing persistent memory, sophisticated task planning, a codebase knowledge graph, and semantic understanding to power intelligent workflows like code review and project management.

Setup Requirements

  • ⚠️Requires Google Gemini API Key (Paid Service)
  • ⚠️Requires Tavily Search API Key (Paid Service)
  • ⚠️Specific client-side configuration needed for full OI OS integration (see OI.md)
Verified SafeView Analysis
The server demonstrates strong security practices including: reading API keys from environment variables, extensive input validation (Ajv schemas) for tool arguments, robust path sanitization and containment checks to prevent path traversal, and the use of prepared statements for SQLite database interactions. While a fully exhaustive audit would require reviewing all dependencies and the complete `DatabaseUtilityService` implementation, the patterns observed in the provided snippets indicate a high level of security awareness. No direct `eval` or uncontrolled command execution from user input was identified.
Updated: 2025-11-23GitHub→
0
0
Low Cost
kazheko icon

db-mcp-server

by kazheko

Sec9

Hosts a Model Context Protocol (MCP) server exposing a stubbed MSSQL query tool for structural data previewing to LLM clients.

Setup Requirements

  • ⚠️Node.js 20.x required
Verified SafeView Analysis
The server explicitly uses a `StubMssqlAdapter` which is designed to never open network sockets or use credentials, mitigating direct database access risks. Input SQL queries are only used for string matching (`includes`) to select templates or simulate errors, not for direct execution. No 'eval' or similar dangerous dynamic code execution patterns are observed. The `StdioServerTransport` implies IPC communication rather than open network ports. The primary dependency `@modelcontextprotocol/sdk` handles core server logic and may involve `cross-spawn` for process management, which is typical for SDKs but not directly controlled by this server's specific implementation with user input.
Updated: 2025-12-01GitHub→
PreviousPage 378 of 760Next