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
alexistomaselli icon

project-central-mcp-server

by alexistomaselli

Sec7

This server provides a set of tools for an AI agent to manage software development projects, including creating and updating projects and issues, listing them, and adding comments, all persisted in a Supabase backend.

Setup Requirements

  • ⚠️Requires a Supabase project with predefined tables (projects, issues, activities, comments) and their respective schemas. These are not automatically created by the server.
  • ⚠️Mandatory environment variables: SUPABASE_URL and SUPABASE_SERVICE_ROLE_KEY must be configured.
  • ⚠️Requires Node.js runtime and project dependencies to be installed (e.g., via npm or yarn).
Verified SafeView Analysis
The server uses a highly privileged `SUPABASE_SERVICE_ROLE_KEY`, which grants extensive access to the Supabase project. Secure management of this environment variable is critical to prevent unauthorized database access in case of server compromise. While the Supabase client library generally prevents SQL injection, directly returning raw Supabase error messages to the client could expose internal database details. CORS is enabled for all origins, which is common but means the API is globally accessible.
Updated: 2026-01-18GitHub→
0
0
Medium Cost
Vishakhavel icon

filesystem-mcp-server

by Vishakhavel

Sec9

Provides an agent or external system with controlled file system access (listing, reading, searching) over a JSON-RPC stdin/stdout interface.

Setup Requirements

  • ⚠️The `MCP_FS_ROOT` environment variable must be explicitly set to the desired root directory, as the default is a user-specific path that is unlikely to be appropriate for general use.
  • ⚠️Requires Node.js and `ts-node` to run directly from source, or requires a prior build step (`tsc`) if running the compiled JavaScript output.
Verified SafeView Analysis
The server implements robust path traversal protection using `resolveSafePath` to confine all file operations strictly within a configured `ROOT_DIR`. File read operations are limited by `MAX_READ_BYTES` to prevent excessive memory usage or denial-of-service, and the search tool crudely filters out common binary file types. Communication occurs via stdin/stdout, which inherently limits direct network attack vectors. The default `ROOT_DIR` (`/Users/vichu/Desktop/`) is user-specific and highlights the critical need for users to configure the `MCP_FS_ROOT` environment variable to a sensible, isolated directory.
Updated: 2025-11-27GitHub→
0
0
Low Cost
BlockRunAI icon

awesome-creator-mcp

by BlockRunAI

Sec10

This repository serves as a curated directory of Model Context Protocol (MCP) servers tailored for content creators, designers, and media professionals, facilitating the discovery of tools for various creative tasks.

Verified SafeView Analysis
The provided source code for `awesome-creator-mcp` consists of Markdown files (README.md, CONTRIBUTING.md) which describe a curated list of external MCP servers. This repository itself does not contain any executable code, server logic, or dynamic elements that could introduce security vulnerabilities such as 'eval', obfuscation, network risks, or hardcoded secrets. It is safe to clone and view. Any security considerations would apply to the individual MCP servers linked within this directory, not to this repository itself.
Updated: 2026-01-18GitHub→
0
0
Medium Cost
BarrosEdu icon
Sec9

Provides an AI-powered server for analyzing earthquake data and seismic risk using a multi-agent system and external APIs.

Setup Requirements

  • ⚠️Requires OpenAI API Key (Paid)
  • ⚠️Requires AGENTOPS_API_KEY (Optional for tracing, but recommended)
  • ⚠️Requires API_KEY for the external earthquake data service
  • ⚠️Requires Python 3.13 or higher
Verified SafeView Analysis
Relies on environment variables for sensitive API keys (OpenAI, AgentOps, custom earthquake API), which is good practice. No 'eval' or malicious obfuscation found. Network calls are made to specified external APIs (earthquake data, OpenAI, AgentOps) and a local sub-MCP server, introducing typical network dependency risks. The nested execution of `mcp/earthquake_mcp_server.py` is controlled and not arbitrary code execution.
Updated: 2025-11-22GitHub→
0
0
Low Cost
Maxymize icon
Sec8

Transforms traditional MCP (Model Context Protocol) servers into efficient Code Execution skills to significantly reduce token consumption for AI agents.

Setup Requirements

  • ⚠️Requires Claude Code CLI
  • ⚠️Requires Node.js 18+
  • ⚠️Requires TypeScript
  • ⚠️OpenAI API Key (Paid) needed for AI-powered/Hybrid skills (e.g., Sentry's AI search, shadcn-vue's requirement analysis)
Verified SafeView Analysis
The system utilizes `child_process.spawn` to launch other MCP servers, which could be a risk if arbitrary commands were injectable; however, the `MCP_SERVERS` configuration in `client.ts` hardcodes known server commands, limiting direct injection. The project's documentation (`CLAUDE.md`) explicitly outlines robust security measures for the code execution environment (e.g., sandbox limits, PII tokenization, restricted filesystem access, network whitelisting, secure environment variables), mitigating risks associated with generated agent code and local execution. Proper management of environment variables is critical.
Updated: 2026-01-08GitHub→
0
0
Low Cost
BapanBigData icon
Sec9

This server provides a set of callable tools, specifically an addition function and a real-time weather retriever, to be consumed by an MCP agent or other clients.

Setup Requirements

  • ⚠️Requires Python 3.12 or higher.
  • ⚠️Requires an OpenWeatherMap API key (OPENWEATHER_API_KEY) to be set as an environment variable for the weather tool to function.
  • ⚠️The server itself expects an '--api-key' argument for its own operation, likely for client authentication.
Verified SafeView Analysis
The server correctly loads API keys from environment variables, avoiding hardcoded secrets. It performs external network requests to OpenWeatherMap, which is standard for fetching external data. No 'eval' or other dynamic code execution patterns were found, enhancing safety. One point deduction for external API dependency.
Updated: 2025-12-06GitHub→
0
0
Low Cost
edwardedmonds icon

sift-releases

by edwardedmonds

Sec6

Provides persistent memory, fast search, and intelligent file editing for Claude Code sessions, enhancing AI developer workflows.

Setup Requirements

  • ⚠️Core 'sift' binary is proprietary and closed-source.
  • ⚠️Requires 'claude' CLI to be installed and configured for MCP server registration and interaction.
  • ⚠️Web crawling (`sift_web_crawl`) requires `libcurl` to be installed (e.g., `libcurl4-openssl-dev` on Debian/Ubuntu, `curl` on macOS).
Verified SafeView Analysis
The core 'sift' binary, which functions as the MCP server, is proprietary and distributed as a binary only. This prevents a full security audit of its C source code. While the Python installer scripts (sift-setup.py, sift-uninstall.py) perform standard privileged operations (file system modifications, subprocess execution) typical for such installations, the inherent trust placed in the closed-source binary is significant. Tools like `sift_sql` or `sift_edit` (with its patch mode) could pose security risks (e.g., SQL injection, arbitrary file manipulation) if inputs from the interacting LLM are not meticulously sanitized and validated within the C implementation. The system design implicitly relies on robust internal handling and careful constraint of the LLM's capabilities to prevent abuse.
Updated: 2026-01-18GitHub→
0
0
Medium Cost
echo-smith664 icon

compound-mcp-server

by echo-smith664

Sec9

Provides a Model Context Protocol (MCP) server that enables AI clients (like Claude Desktop) to interact with Groq's compound/meta models, offering tools for real-time information retrieval and simulated code execution.

Setup Requirements

  • ⚠️Requires a Groq API key (GROQ_API_KEY) for basic functionality.
  • ⚠️Node.js >= 18.0.0 is required.
  • ⚠️For Vercel deployments, a Redis instance (REDIS_URL) is recommended for SSE transport.
Verified SafeView Analysis
The server's core functionality involves proxying user questions to the Groq API. It correctly uses environment variables for sensitive API keys (GROQ_API_KEY) and Redis connection strings (REDIS_URL), preventing hardcoding of secrets. There is no direct `eval` or arbitrary code execution by the server based on user input. The 'code execution' tool name refers to the *model's* capability, not the server's. The use of standard, well-maintained SDKs (`@modelcontextprotocol/sdk`, `groq-sdk`, `@vercel/mcp-adapter`) suggests good security practices in dependencies.
Updated: 2025-11-20GitHub→
0
0
Low Cost
carlosduplar icon

multi-agent-mcp

by carlosduplar

Sec8

Intelligent routing and guidance layer for AI coding agents, enabling seamless delegation of specialized tasks to various tools while maintaining context and privacy.

Setup Requirements

  • ⚠️Requires Python 3.10+
  • ⚠️Requires at least one external AI agent CLI installed (e.g., Gemini CLI via npm, Aider via pip, Claude Code via npm)
  • ⚠️Installer automatically configures Claude Code, which might require specific Claude CLI versions or local setup.
Verified SafeView Analysis
The core `DelegationMCPServer` (when run as `delegation-mcp`) is designed as a routing guidance layer and explicitly states 'No Code Execution' for user tasks. It only suggests commands to the client. It uses `subprocess` for agent discovery (`which`, `--version` checks), which is low-risk. The `OrchestratorRegistry.execute` method, used by other components like `DelegationEngine` and `WorkflowEngine` (e.g., via `delegation-workflow` CLI), *does* execute external commands. This method constructs `subprocess.create_subprocess_exec` commands using a list of arguments, which is generally safer against shell injection than a single string. It also sanitizes the environment passed to subprocesses using an allowlist approach. The `WorkflowContext.interpolate` method includes a warning about handling escaping for interpolated strings passed to shell commands, indicating a potential point of concern if the external CLIs themselves have argument parsing vulnerabilities or if workflow inputs are untrusted. No direct shell injection vulnerabilities were found within the project's subprocess calls, but reliance on external CLIs means their security model is inherited.
Updated: 2025-11-28GitHub→
0
0
Low Cost
michaelpetrik icon

mcp-setter

by michaelpetrik

Sec8

Installing and managing Model Context Protocol (MCP) servers across different AI services and operating systems.

Setup Requirements

  • ⚠️Requires Node.js 18+ LTS and npm 9+ or pnpm.
  • ⚠️Requires a `.env` file (based on `.env.example`) to be configured with API keys for specific MCP servers (e.g., CONTEXT7_API_KEY, GITHUB_MCP_PAT, N8N_API_KEY, N8N_API_URL) for full functionality of the MCP servers it manages, though the tool itself can run without them.
  • ⚠️The `requestElevation()` permissions function is a placeholder and does not implement actual OS-specific privilege elevation, requiring manual elevation for operations that need it.
Verified SafeView Analysis
The codebase demonstrates strong security practices including explicit handling of sensitive environment variables, runtime validation, and scrubbing of sensitive data (API keys, file paths) from telemetry payloads. It also includes business rule validation to prevent potentially dangerous system commands from being executed. However, its core functionality involves installing and executing code from an external MCP registry, which inherently carries a supply chain risk dependent on the trustworthiness of the registry's content. No 'eval' or malicious obfuscation patterns were found.
Updated: 2025-11-19GitHub→
0
0
Medium Cost
Bin4yi icon
Sec8

An AI-powered pet care chatbot system integrating a secured MCP server with an intelligent LangGraph agent, using Asgardeo for authentication and OpenAI for AI capabilities.

Setup Requirements

  • ⚠️Requires Python 3.12 or higher.
  • ⚠️Requires an Asgardeo account and manual application setup (Client ID, Tenant Name, JWKS URL).
  • ⚠️Requires an OpenAI API key (paid service) for AI-powered features.
Verified SafeView Analysis
The project uses robust OAuth2/OIDC authentication with Asgardeo and JWT token validation via JWKS, which are good security practices. Environment variables are correctly used for credentials, with warnings against committing them. However, the default CORS policy (`allow_credentials=True, expose_headers='*', allow_headers='*', allow_methods='*'`) is very permissive and should be restricted to specific origins for a production deployment. The `ssl_verify=False` option for development is mentioned in comments, which, while not active by default, points to a potential development pattern that should be avoided in production.
Updated: 2025-12-02GitHub→
0
0
Low Cost
eladjak icon
Sec8

A central registry for discovering, listing, and recommending AI development skills based on their local markdown files.

Setup Requirements

  • ⚠️Requires `npm install` and `npm run build` within its directory before execution.
Verified SafeView Analysis
The server operates locally on skill markdown files (`SKILL.md`) found in `~/.claude/skills/` and `.claude/skills/`. It uses `YAML.parse` to extract metadata from frontmatter. While `YAML.parse` can pose risks with untrusted input, the server is designed to process local, presumably trusted skill definitions. No direct network risks or hardcoded secrets were identified within the server's own logic. The primary security concern would be if malicious `SKILL.md` files could be injected into these paths, potentially leading to local privilege escalation or unexpected behavior if the YAML parsing or subsequent markdown rendering (by the consuming AI/tool) is vulnerable. However, the server itself does not expose external network interfaces and focuses on local file processing.
Updated: 2026-01-18GitHub→
PreviousPage 564 of 760Next