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(8554)

0
0
Low Cost
tingz-personal icon

mcp-tool-tingz

by tingz-personal

Sec7

A Python toolkit for building Model Context Protocol (MCP) servers, allowing exposure of custom tools via JSON-RPC 2.0 over WebSocket or stdio.

Setup Requirements

  • ⚠️Requires Python 3.12 or newer.
  • ⚠️Relies on standard Python packages (FastAPI, uvicorn, typer, pydantic) which are easily installed but must be present.
Verified SafeView Analysis
The server leverages FastAPI for WebSocket endpoints and uses `asyncio` for stdio transport. JSON-RPC requests are deserialized and dispatched to registered tool handlers. The `import_app` utility can load arbitrary Python modules/files, which, while intended for tool registration, could be a security risk if the `--app` argument is controlled by an untrusted source. Tool handlers receive arguments directly from JSON-RPC requests (`**arguments`), requiring developers to implement robust input validation within their custom tools to prevent injection or unexpected behavior. No obvious hardcoded secrets or malicious patterns were found in the provided code, but the powerful dynamic code loading and direct argument passing necessitate careful usage and deployment practices by the developer.
Updated: 2025-11-26GitHub→
0
0
Low Cost
michaelahern icon

airthings-consumer-mcp

by michaelahern

Sec9

Provides real-time air quality data from Airthings devices to AI assistants via the Model Context Protocol (MCP).

Setup Requirements

  • ⚠️Requires an Airthings Developer Account to obtain 'client_id' and 'client_secret' for API access.
  • ⚠️Requires Node.js version 20 or higher.
  • ⚠️An MCP client (e.g., Claude Desktop, Amazon Q, 'mcpb' CLI) is needed to interact with the server.
Verified SafeView Analysis
Secrets (Airthings Client ID and Secret) are properly handled via environment variables and are not hardcoded, adhering to secure coding practices. There is no usage of 'eval' or other dangerous dynamic code execution patterns. The server's core function is to act as a proxy to a third-party Airthings API via the 'airthings-consumer-api' library; therefore, external risks would primarily depend on the security and trustworthiness of that external API and library.
Updated: 2026-01-17GitHub→
0
0
Medium Cost
Microck icon
Sec5

Automate Discord user actions and interactions through an AI agent, leveraging a wide range of tools for comprehensive user autonomy.

Setup Requirements

  • ⚠️Using a Discord selfbot violates Discord's Terms of Service and can lead to permanent account termination.
  • ⚠️Initial setup requires extracting a Discord user token either manually or via an automated browser process, which demands user interaction and trust in the setup script.
  • ⚠️Automated captcha solving (for joining servers) is optional but requires an API key for a paid third-party service (e.g., CapSolver, CapMonster, NopeCHA).
Review RequiredView Analysis
The primary security concern is the inherent nature of a 'selfbot', which violates Discord's Terms of Service and carries a significant risk of account termination. The setup process involves handling a Discord user token, which grants full access to the account, requiring careful management by the user. Optional integration with third-party captcha-solving services (CapSolver, CapMonster, NopeCHA) means sending captcha-related data to external APIs, which is a privacy consideration. The `download_attachment` tool allows downloading files to the local filesystem, which could be a vector for malicious files if the AI is instructed to download untrusted content to sensitive locations. The code base itself uses Zod for input validation and has a `DANGER_MODE` setting for sensitive relationship actions, indicating an attempt at internal safety, but these do not mitigate the fundamental risks of selfbotting or user-directed malicious actions.
Updated: 2026-01-17GitHub→
0
0
Low Cost
Yutaka-Murata icon
Sec8

Real-time monitoring and analysis of Claude Code token usage and cost for local development sessions.

Setup Requirements

  • ⚠️This project is explicitly noted as archived and deprecated in favor of an MCP server version. It has known issues, specifically incorrect `cache_read` pricing ($1.50/MTok vs correct $0.50/MTok) which is corrected in the MCP version.
  • ⚠️Requires Python 3 to be installed and available in the system's PATH.
  • ⚠️Full functionality, including desktop notifications for alerts, requires macOS due to the use of `osascript`.
  • ⚠️Assumes a specific local directory structure for Claude Code logs and transcripts (`~/.claude/token_logs` and `~/.claude/projects`).
Verified SafeView Analysis
The scripts utilize `python3 -c "..."` to embed shell variables (e.g., log entries, pricing rates) into Python strings for processing JSON. While this pattern carries an inherent, albeit low, risk of command injection if the source data (Claude Code transcript files) were malicious or improperly formed, it's generally considered safe within a controlled local environment where Claude Code itself is assumed to be trusted. No direct `eval` commands are used in Bash. No hardcoded credentials or external network risks are identified. macOS notifications via `osascript` use extracted log data, posing a minimal risk if log content were specifically crafted to exploit shell commands.
Updated: 2026-01-17GitHub→
0
0
Low Cost
theshivamlko icon

flutter-ai-labs

by theshivamlko

Sec3

A Dart-based server implementing the Model Context Protocol (MCP) to enable AI agents to perform email sending operations.

Setup Requirements

  • ⚠️Requires a Dart runtime environment to be installed.
  • ⚠️Requires external SMTP service credentials (username, password, server address, port) which are hardcoded in the example and must be replaced with actual secure credentials.
  • ⚠️Initial setup involves running `dart pub get` to install dependencies within the `email_dart_mcp_server_example` directory.
  • ⚠️Integration with specific IDEs (like Cursor) or the MCP Inspector requires manual configuration of the server's exact file path on the user's system.
Review RequiredView Analysis
The example code explicitly hardcodes SMTP server credentials (username and password) directly into the source file for email sending. While the README advises to replace these values, the presence of hardcoded credentials in the example code itself is a significant security risk if deployed or committed as-is, potentially leading to accidental exposure of sensitive information.
Updated: 2025-12-01GitHub→
0
0
Medium Cost
caioldcarvalho icon

gtm-mcp

by caioldcarvalho

Sec10

Automates Google Tag Manager operations and deployments using natural language commands via Claude AI.

Setup Requirements

  • ⚠️Requires Claude AI access/API Key (Paid)
  • ⚠️Requires Google Tag Manager API access and proper authentication/permissions
  • ⚠️Installation and usage instructions are currently missing from the README
Verified SafeView Analysis
The provided 'SOURCE CODE' is only the `README.md` file. Based solely on the contents of the README, no security vulnerabilities, malicious patterns, or dangerous functions (like `eval`) can be identified. A comprehensive security audit would require access to the actual application source code files.
Updated: 2026-01-16GitHub→
0
0
Medium Cost

SpreadJS_mcp

by CharlieNey

Sec3

A web application with an AI chat interface alongside a spreadsheet, allowing users to interact with the spreadsheet through natural language using the SpreadJS MCP server to execute operations.

Setup Requirements

  • ⚠️Requires ANTHROPIC_API_KEY (a paid service) for AI chat functionality.
  • ⚠️Requires a Supabase project (SUPABASE_URL, SUPABASE_ANON_KEY) for file storage and management.
  • ⚠️SpreadJS license (VITE_SPREADJS_LICENSE or SPREADJS_LICENSE_KEY) is optional but needed for the full Excel-like UI and advanced features like Excel I/O.
  • ⚠️The `SPREADJS_MCP_PATH` environment variable needs to point correctly to the built MCP server, typically `../../dist/index.js` when run from the `packages/backend` directory.
Review RequiredView Analysis
The `src/tools/io.ts` module allows the AI (driven by user prompts) to read from and write to the local filesystem via tools like `import_json`, `export_json`, `import_csv`, and `export_csv` using a user-provided `filePath`. This poses a critical security risk, as a malicious prompt could instruct the AI to read sensitive server files (e.g., `/etc/passwd`) or write arbitrary files, potentially leading to remote code execution or data exfiltration. Additionally, the backend spawns the MCP server as a child process; while the default path is relative, if `SPREADJS_MCP_PATH` were to be controlled by an attacker, it could lead to arbitrary code execution. The application also handles arbitrary file uploads, which are then analyzed by the AI, increasing the potential attack surface if combined with the filesystem manipulation tools.
Updated: 2025-11-26GitHub→
0
0
Medium Cost
major icon

pcp-mcp

by major

Sec9

Provides an MCP server to query real-time system performance metrics from Performance Co-Pilot (PCP), including CPU, memory, disk I/O, network, and processes, enabling LLM-driven system analysis and troubleshooting.

Setup Requirements

  • ⚠️Requires Performance Co-Pilot (pmcd and pmproxy daemons) to be installed and actively running on the target system(s).
  • ⚠️Requires Python 3.10 or newer.
  • ⚠️For monitoring remote systems, the PCP_TARGET_HOST environment variable must be set. If tools are used to query arbitrary hosts via their 'host' parameter, PCP_ALLOWED_HOSTS must be explicitly configured (e.g., to ['*'] or a specific list of hostspecs) for such requests to be permitted.
Verified SafeView Analysis
The server correctly uses environment variables for sensitive credentials (PCP_USERNAME, PCP_PASSWORD) and enables TLS verification by default (PCP_TLS_VERIFY). It implements a robust allowlist mechanism (PCP_ALLOWED_HOSTS) to restrict which remote hosts can be queried via the 'host' parameter in tools, mitigating Server-Side Request Forgery (SSRF) risks. No 'eval', 'exec', or other dynamic code execution patterns were found. The implementation demonstrates good security practices for a network-facing service.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
gesteves icon

domestique

by gesteves

Sec9

Integrates fitness data from Intervals.icu, Whoop, and TrainerRoad into a unified MCP server for comprehensive analysis and workout management via AI agents.

Setup Requirements

  • ⚠️Node.js 20+ is a prerequisite.
  • ⚠️Intervals.icu API Key and Athlete ID are mandatory for core functionality.
  • ⚠️Whoop integration requires a multi-step OAuth authorization process involving manual browser interaction and a Redis database for token storage and refresh management. Redis is a hard dependency for Whoop functionality.
  • ⚠️TrainerRoad integration requires a private iCal calendar feed URL.
Verified SafeView Analysis
The project demonstrates strong security practices including the use of environment variables for all sensitive credentials, secure constant-time comparison for the MCP authentication token, and all external API communications use HTTPS. The Whoop OAuth token refresh mechanism is well-implemented with Redis-based distributed locking to prevent race conditions and token invalidation. Error handling is structured to avoid exposing raw details to LLMs. No 'eval' or obfuscation found.
Updated: 2026-01-17GitHub→
0
0
Medium Cost
sundar-nallalagappan icon

FastMCP_Samples

by sundar-nallalagappan

Sec8

Orchestrating multi-protocol communication protocol (MCP) tool servers with various LLM agent frameworks for diverse task execution.

Setup Requirements

  • ⚠️Requires the 'weatherserver.py' script to be run separately in the background for HTTP connections.
  • ⚠️Requires API keys (e.g., GROQ_API_KEY, OPENAI_API_KEY) for LLM providers, which typically involve paid services.
  • ⚠️Python 3.11 or higher is required as specified in 'pyproject.toml'.
Verified SafeView Analysis
The code uses environment variables for API keys (e.g., GROQ_API_KEY, OPENAI_API_KEY), which is good practice. No 'eval' or obvious malicious patterns were found. The 'weatherserver.py' runs an HTTP server on localhost, which is typical for local development/testing and not inherently insecure when confined to localhost. 'mathserver.py' uses stdio transport, which is generally secure. Observability fixes in ADK clients indicate attention to robustness.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
ricCap icon
Sec9

This server integrates Sketch Engine's corpus linguistics and text analysis tools with Claude AI, allowing users to perform complex linguistic research and language analysis directly through conversational prompts.

Setup Requirements

  • ⚠️Requires a free Sketch Engine account to obtain an API key.
  • ⚠️Requires manual download and execution of a standalone executable specific to the user's OS.
  • ⚠️Requires manual editing of a specific Claude Desktop JSON configuration file to register the server and API key.
  • ⚠️Requires Node.js >= 21.0.0 if building/running from source.
Verified SafeView Analysis
No critical vulnerabilities found. The Sketch Engine API key is correctly handled via environment variables and is not hardcoded. Input validation is performed using Zod schemas for all tool calls, mitigating potential injection risks. Local file system writes are constrained to a '.cache' directory within the application's current working directory, which is appropriate for a local agent.
Updated: 2025-12-14GitHub→
0
0
Medium Cost
guvensoft icon

mcp-rag-server

by guvensoft

Sec3

Provides a local Model Context Protocol (MCP) server that analyzes codebases and offers context-aware search and development tools to AI agents.

Setup Requirements

  • ⚠️Python 3.9+ is required, and for the full-featured semantic engine (Sentence-Transformers + ChromaDB), additional Python packages (`sentence-transformers`, `chromadb`, `fastapi`, `uvicorn`) must be manually installed via pip. Without these, a less capable Node.js fallback engine is used.
  • ⚠️Redis (e.g., Memurai on Windows) is implicitly recommended for the BullMQ job queue and QueryCache, adding an external dependency for enhanced indexing and performance.
  • ⚠️The initial indexing process can be time-consuming for large codebases, though incremental indexing and a watcher are in place to manage updates.
Review RequiredView Analysis
The server includes tools like `run_tests` and `run_task` that execute arbitrary shell commands or npm scripts based on agent input. While the project documentation mentions sandbox modes and approval policies for clients, the server itself allows direct execution of these commands. This represents a significant security risk if the server is exposed to untrusted inputs or not operated within a robustly sandboxed environment. File access is somewhat restricted by a policy layer (blocking sensitive file types and large files, limiting to specified roots), and network services bind to localhost by default, mitigating external attack surface.
Updated: 2025-11-25GitHub→
PreviousPage 377 of 713Next