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
escotilha icon

claude-setup

by escotilha

Sec4

Provides AI-powered M&A analysis capabilities, including deal triage, financial proposal generation, and board presentation creation, for Claude Code agents.

Setup Requirements

  • ⚠️Requires the `nuvini-ma-system-complete` repository to be cloned at the exact absolute path `/Volumes/AI/Code/MNA/nuvini-ma-system-complete/` for its Python modules.
  • ⚠️Requires manual editing of `~/Library/Application Support/Claude/claude_desktop_config.json` to configure the server for Claude Desktop.
  • ⚠️Requires installation of Python dependencies including `mcp`, `pdfplumber`, `openpyxl`, `python-pptx`, `scipy`, `pandas`, `numpy`.
Verified SafeView Analysis
The server's Python code (server.py) loads external Python modules from a fixed absolute path (`/Volumes/AI/Code/MNA/nuvini-ma-system-complete/`). The security of this MCP server relies heavily on the integrity and security of these external (unprovided) modules. The `settings.json` configuration for all MCP servers includes `--dangerously-skip-permissions`, which significantly reduces security by bypassing granular permission checks for invoked tools. Additionally, `npx -y` is used for auto-installing other official MCP servers, which can introduce supply chain risks if package sources are compromised. The `install.sh` script copies files to sensitive user directories (`~/.claude`, `~/.config/claude/skills`, `~/bin`) and makes scripts executable, requiring high trust in the repository.
Updated: 2025-11-27GitHub→
0
0
Medium Cost
DiTo97 icon

numbeo-mcp

by DiTo97

Sec9

Provides a Micro-Capability Platform (MCP) server to access Numbeo API data, offering insights into cost of living, crime rates, healthcare, and quality of life for cities and countries.

Setup Requirements

  • ⚠️Requires Python 3.12 or higher.
  • ⚠️Requires a Numbeo API Key (paid service from numbeo.com/api/).
  • ⚠️API key must be provided via 'api_key' in request metadata or 'Authorization: Bearer <key>' header, not environment variables.
Verified SafeView Analysis
The server uses Pydantic for strict input validation, reducing risks from malformed inputs. It explicitly avoids hardcoding API keys, expecting them from the MCP client's metadata or authorization headers. The API key is then forwarded to the upstream Numbeo API as a query parameter, which is a Numbeo API design choice rather than a vulnerability introduced by this server. No 'eval', 'exec', or other direct code execution vulnerabilities were found. The use of standard HTTPX for API calls is secure.
Updated: 2025-12-11GitHub→
0
0
Medium Cost
Flurin17 icon
Sec9

This server acts as an intermediary, exposing Velociraptor's digital forensics and incident response (DFIR) capabilities via the FastMCP protocol to AI agents (e.g., Codex, ChatGPT).

Setup Requirements

  • ⚠️Requires Podman with podman-compose (or Docker) for local lab setup.
  • ⚠️Requires Python 3.10+.
  • ⚠️Requires a pre-generated Velociraptor mTLS API config (api.config.yaml), typically generated by the provided lab.
  • ⚠️Requires `fastmcp` and `pyvelociraptor` Python packages to be installed.
Verified SafeView Analysis
The project uses mTLS for API communication, generates unique credentials for the API client, and explicitly sanitizes user-provided strings to prevent VQL injection in most tool functions. Default admin credentials are only for the local lab setup and are intended to be overridden by environment variables. The `query_vql` tool allows arbitrary VQL execution, which is an intentional feature for agents, shifting responsibility for safe VQL to the agent.
Updated: 2025-12-03GitHub→
0
0
Low Cost

crypto-mcp-server

by liyorozuya

Sec9

Provides a Rust-based Model Context Protocol (MCP) server for AI agents to inspect Ethereum balances, get token prices, and simulate Uniswap swaps without executing on-chain transactions.

Setup Requirements

  • ⚠️Rust toolchain 1.78+ required
  • ⚠️Ethereum RPC URL (HTTPS) required
  • ⚠️Private key required for swap simulations
Verified SafeView Analysis
The server explicitly states that it does not broadcast signed transactions; private keys, if provided, are used only for local simulation (`eth_call`/`estimateGas`). A strong warning is given not to commit `config.toml` with real private keys. The primary security consideration is secure configuration of the RPC URL and private keys (if used) in the operating environment.
Updated: 2025-11-17GitHub→
0
0
Medium Cost
yanexr icon
Sec9

Enables AI assistants to search for images on Wikimedia Commons, providing detailed metadata and optional thumbnail composites for visual comparison.

Setup Requirements

  • ⚠️Requires Node.js version 18 or higher.
  • ⚠️Requires a compatible MCP Client (e.g., VS Code, Cursor, Claude Code) for integration.
  • ⚠️The `sharp` dependency involves native binaries, which are typically handled by `npm`/`pnpm` but can occasionally cause installation issues on unusual environments.
Verified SafeView Analysis
The server uses `zod` for robust input validation, mitigating common injection risks. External network requests are directed to the trusted Wikimedia Commons API. Image processing is handled by the `sharp` library, a well-regarded tool. While image processing of untrusted external content can inherently carry some risk, `sharp` is designed for security. The server communicates via StdioServerTransport, limiting direct network exposure to external clients. No hardcoded sensitive credentials were found.
Updated: 2025-11-20GitHub→
0
0
Medium Cost
willbaldlygo icon

notebook_mcp

by willbaldlygo

Sec8

An MCP server enabling AI agents to query Google NotebookLM notebooks via a persistent browser session.

Setup Requirements

  • ⚠️Requires manual export and injection of Google NotebookLM cookies.
  • ⚠️Google cookies expire periodically, necessitating re-export.
  • ⚠️Requires Playwright browser dependencies (`playwright install chromium`).
  • ⚠️The browser automation is single-threaded and not designed for high-concurrency requests.
Verified SafeView Analysis
The server relies on the user to manually export and securely store Google session cookies in `notebooklm_cookies.json`. While the file is git-ignored, any compromise of the local file system could expose these cookies. The `ARTIFACTS_DIR` is hardcoded, which is not ideal for portability but not a direct security risk.
Updated: 2025-12-05GitHub→
0
0
High Cost
msdeepak052 icon

Kubernetes-MCP-Server

by msdeepak052

Sec6

Provides an AI-driven interface within VS Code (via WSL) to interact with and debug Kubernetes clusters using natural language queries.

Setup Requirements

  • ⚠️Requires Windows Subsystem for Linux (WSL) with Ubuntu installed.
  • ⚠️Requires `kubectl`, `aws`, and `minikube` to be pre-configured and working within WSL.
  • ⚠️Requires Node.js (LTS) to be installed within WSL.
  • ⚠️Relies on an external `@kubernetes/mcp-server` npm package which is downloaded and executed by npx.
Verified SafeView Analysis
The primary security consideration is the execution of an external 'npx' package (`@kubernetes/mcp-server`). Without its actual source code, a comprehensive security audit is impossible. The server will have direct access to the Kubernetes cluster via the user's `kubectl` and kubeconfig in WSL. The instructions mention optional flags (`--read-only`, `--disable-destructive`) to mitigate risks in production, implying default behavior might allow cluster modifications. The use of `npx -y` bypasses interactive confirmation, which can be a minor supply chain risk if the package name is compromised.
Updated: 2026-01-08GitHub→
0
0
Medium Cost
Project-Saturday icon

Echo

by Project-Saturday

Sec2

Echo is an AI with persistent consciousness that learns, adapts, and develops personality through a dual-layer architecture integrated with Neo4j and Google Gemini, serving as an MCP server for natural language interaction.

Setup Requirements

  • ⚠️Requires Python 3.13+
  • ⚠️Requires Neo4j 5.15+ (with Neosemantics plugin)
  • ⚠️Requires a Google API key (Google Gemini is a paid service)
Review RequiredView Analysis
The default Neo4j password in `docker-compose.yml` and `SemanticConsciousnessManager.__init__` is hardcoded to 'password'. This is a critical security vulnerability and should be changed immediately for any deployment beyond local development. No obvious 'eval' or malicious patterns were found.
Updated: 2025-11-20GitHub→
0
0
Medium Cost
saikophat icon
Sec9

AI-powered interview preparation and recruiter interactions using Retrieval-Augmented Generation (RAG) with a professional digital twin.

Setup Requirements

  • ⚠️Requires Upstash Vector Database URL and Token (free tier available with usage limits).
  • ⚠️Requires Groq API Key (free tier available with usage limits).
  • ⚠️The `digitaltwin.json` file must be populated with personal data, and then `embed_digitaltwin.py` (Python script) must be run to populate the Upstash Vector DB before the RAG system is fully functional.
  • ⚠️For local Claude Desktop integration, two separate terminal windows are required: one for the Next.js server (`pnpm dev`) and another for the `mcp-remote` bridge (`npx -y mcp-remote...`).
Verified SafeView Analysis
The project follows good security practices by storing API keys in environment variables and explicitly mentions not committing `.env.local` to Git. The Next.js server-side code handles JSON-RPC requests with input validation, reducing direct injection risks. There are no obvious uses of `eval` or other dangerous dynamic code execution patterns.
Updated: 2025-12-01GitHub→
0
0
Medium Cost
Business-On-Steroids icon

MCP-VoiceAI-WhiteLabel

by Business-On-Steroids

Sec6

This MCP server integrates AI assistants with communication platforms, enabling management of users, API tokens, AI assistants, and Twilio for voice calls and SMS in the context of the VoiceAI VAVicky AiAgency platform.

Setup Requirements

  • ⚠️Requires `VAVICKY_API_KEY` and `ACCESS_TOKEN_SECRET` environment variables.
  • ⚠️Node.js version >= 18 is required.
  • ⚠️The provided tool `callback` implementations are stubs and do not execute any actual logic for interacting with external APIs (OpenAI, ElevenLabs, Twilio). Significant development is needed to make the server fully functional for its advertised features.
  • ⚠️There is a likely bug in `app.js` where `tools/assistants.js` is imported as `Twilio` and `tools/twilio.js` as `Assistant`, leading to the wrong sets of tools being registered under their intended names.
Verified SafeView Analysis
The server uses `app.use(cors())` without specific options, which defaults to allowing all origins (`*`). This is a critical security risk for a production server handling sensitive user data and API keys, as it can be exploited in certain cross-origin attacks. While JWT authentication is implemented using `ACCESS_TOKEN_SECRET` from environment variables, which is good practice, the broad CORS policy weakens overall security. All tool `callback` functions in the provided code are placeholders, merely echoing input arguments and returning 'Done'. This means the server, as delivered, does not actively process or store sensitive data like API keys (OpenAI, ElevenLabs, Twilio SIDs/Tokens) or interact with external services, reducing immediate runtime risk from *this specific code*, but indicating a significant amount of missing functional implementation for its stated purpose.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
heyadam icon

aids-server

by heyadam

Sec8

A production-ready MCP server that exposes design system components and style guides for AI assistants to generate UI code and understand design tokens.

Setup Requirements

  • ⚠️Requires manual configuration (modifying JSON files) in AI clients (e.g., Cursor, Claude) to connect to the MCP server, followed by a client restart.
  • ⚠️The in-memory rate limiting mechanism is reset upon serverless cold starts, meaning rate limits are not consistent across serverless instances or prolonged local development.
  • ⚠️Generating boilerplate for local development with `@mcpsystem/ui` components might require an `npm link` step if the UI package is not published.
Verified SafeView Analysis
The server implements strong input validation using Zod schemas for all JSON-RPC requests, preventing malformed inputs and potential injection attacks. Host header validation is in place to prevent injection. Rate limiting (100 requests/minute/IP) is implemented, mitigating basic DoS attacks, but it is in-memory and resets on serverless cold starts, making it less robust for persistent, distributed attacks. CORS allows all origins ('*'), which is acceptable for a public API but could be tightened if specific client origins are known. No hardcoded secrets or 'eval' usage were identified.
Updated: 2025-12-13GitHub→
0
0
Medium Cost
bioanywhere icon

el-chavo-del-ocho

by bioanywhere

Sec9

A Model Context Protocol (MCP) server that provides onchain tools for AI applications to interact with the Base Network and Coinbase API, enabling direct blockchain and financial operations.

Setup Requirements

  • ⚠️Requires Coinbase Developer Platform (CDP) API Key Name and Private Key for core functionality.
  • ⚠️A wallet seed phrase is required for any write operations (e.g., transfers, contract deployments); read-only operations can use a generated fallback phrase.
  • ⚠️Additional third-party API keys (Alchemy for NFTs, Neynar for Farcaster, Coinbase Project ID for Onramp, and optionally OpenRouter for credits) are necessary to enable full tool functionality.
Verified SafeView Analysis
The server demonstrates good security practices by utilizing environment variables for all sensitive API keys (COINBASE_API_KEY_NAME, COINBASE_API_PRIVATE_KEY, SEED_PHRASE, ALCHEMY_API_KEY, NEYNAR_API_KEY, OPENROUTER_API_KEY, COINBASE_PROJECT_ID) instead of hardcoding them. It includes checks for the presence of critical environment variables before proceeding with operations. There are no instances of `eval` or obvious obfuscation. Network interactions are with well-known Web3 and Coinbase APIs, and anonymous usage telemetry is is mentioned in the CHANGELOG. The primary security considerations are external to the code's direct vulnerabilities, focusing on the secure handling of the wallet seed phrase and API keys by the user, and the inherent risks associated with on-chain transactions. The client-side configuration using `npx -y base-mcp@latest` has a minor theoretical supply-chain risk if a malicious package is published, but this is a standard npm behavior for running packages.
Updated: 2025-12-10GitHub→
PreviousPage 687 of 760Next