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
Low Cost
spanwich icon

academic-mcp

by spanwich

Sec8

Zotero-integrated server for academic paper analysis, providing chunked access and LLM-powered extractions to overcome large document context limitations, enabling semantic search and on-demand content retrieval.

Setup Requirements

  • ⚠️Requires Zotero with papers and the Better BibTeX plugin installed for proper integration and citation key generation.
  • ⚠️Requires Ollama to be installed and running locally, with the specified LLM model ('qwen2.5:3b' by default) pulled and available.
  • ⚠️The Zotero application must be closed during database import/sync operations to prevent database locking issues (as explicitly handled by the ZoteroReader).
Verified SafeView Analysis
The server interacts with a local Ollama instance for LLM operations, limiting direct exposure to external network threats. It uses `subprocess` calls to manage the Ollama server (start, pull models), which is a common and necessary pattern for integrating local services. Zotero database access is read-only and operates on temporary copies to prevent locking issues and ensure data integrity. Custom JSON parsing logic (`_repair_json`, `_safe_json_loads`) is implemented to robustly handle potentially malformed JSON output from the LLM, reducing the risk of parsing errors. No direct use of `eval()` or hardcoded sensitive credentials beyond the local Ollama host is found. The overall architecture focuses on local processing and read-only access to external data sources, contributing to a relatively high security posture within its operational scope.
Updated: 2026-01-18GitHub→
0
0
High Cost
ashen-dusk icon

mcp-hub

by ashen-dusk

Sec5

Manages, connects to, and interacts with Model Context Protocol (MCP) servers using AI agents and a GraphQL API, providing category-based organization and user-isolated state.

Setup Requirements

  • ⚠️Requires Python 3.12+.
  • ⚠️Requires a running Redis server.
  • ⚠️Requires various API keys (OpenAI/DeepSeek, Tavily, Google Client ID, OpenRouter) depending on features used.
Review RequiredView Analysis
The server allows configuration of `stdio` transport type for MCP servers, which executes arbitrary commands and arguments directly on the host machine. If an authenticated user (even an owner, not just admin) can configure `command` and `args` in the `MCPServer` model, it creates a severe remote code execution vulnerability. There is no explicit input sanitization or whitelisting for these fields. Additionally, the system makes dynamic external requests to `server.url` during OAuth discovery, token exchange, and FastMCP client operations, which could potentially expose the backend to SSRF or other network-based attacks if `server.url` is controlled by a malicious actor and not adequately validated beyond basic URL parsing.
Updated: 2026-01-18GitHub→
0
0
Low Cost
xztaityozx icon
Sec7

Provides a Multi-Client Protocol (MCP) server for the A.I.VOICE Editor to facilitate communication and control.

Setup Requirements

  • ⚠️Requires A.I.VOICE Editor (version 1); it is not compatible with A.I. VOICE Editor 2.
  • ⚠️Functionality is limited to what is publicly exposed by the A.I.Talk.Editor.Api.
Verified SafeView Analysis
The provided source code is limited to the README.md, so a comprehensive security audit of the server's implementation is not possible. No explicit 'eval', obfuscation, hardcoded secrets, or malicious patterns are visible in the truncated content. Actual server code would need to be reviewed for network vulnerabilities, input sanitization, and access control.
Updated: 2025-12-13GitHub→
0
0
Low Cost
Nicolas-Gong icon

api-request-mcp-server

by Nicolas-Gong

Sec8

Sends HTTP/HTTPS API requests, validates JSON responses, and supports proxy configurations.

Setup Requirements

  • ⚠️Requires manual configuration in a specific, non-standard Cline (Windows) path for MCP servers.
  • ⚠️Requires TypeScript compilation (`npm run build`) before it can be run.
  • ⚠️The MCP configuration needs an absolute path to the compiled 'build/index.js' file.
Verified SafeView Analysis
The server's primary function is to make arbitrary network requests based on user input. While the code itself uses standard libraries (Axios) and includes basic input type validation, it does not implement deep content validation for URLs or headers, which could pose a risk (e.g., SSRF) if exposed to untrusted clients without additional safeguards. No direct 'eval', hardcoded secrets, or obvious malicious patterns were found within the provided source. Proxy handling respects environment variables and NO_PROXY settings correctly.
Updated: 2025-12-03GitHub→
0
0
Medium Cost
polysolmcp icon

polysolmcp

by polysolmcp

Sec7

Enables AI agents to interact with Polymarket prediction markets by providing market data, prices, and historical information.

Setup Requirements

  • ⚠️Requires Python 3.10 or higher.
  • ⚠️Requires Polymarket API Key (private key) and Funder Wallet Address, which are sensitive credentials.
  • ⚠️Requires manual configuration in MCP-compatible clients like Claude Desktop, including the absolute path to the repository.
Verified SafeView Analysis
The server requires a Polymarket API private key (`KEY`) and funder wallet address (`FUNDER`) as environment variables. While loading secrets from `.env` is a standard practice, handling a private key inherently carries significant security risk. Users must ensure their local environment and `.env` file are extremely secure to prevent unauthorized access to their Polymarket account. The server communicates via standard input/output (stdio), which limits network exposure of these credentials during operation.
Updated: 2025-12-09GitHub→
0
0
Medium Cost
support-and-care-labs icon

maven-mcps

by support-and-care-labs

Sec9

A collection of Model Context Protocol (MCP) servers designed to aid in Apache Maven improvements by providing LLM-accessible tools and context.

Setup Requirements

  • ⚠️Docker (Docker Compose V2) is required for running the MCP servers.
  • ⚠️The project is an umbrella for multiple independent MCP servers; setup involves navigating to specific subdirectories and following their individual README.adoc files.
  • ⚠️Building documentation locally requires Antora, which is a Node.js/npm application.
  • ⚠️The `mail-mcp` server uses Python scripts for data retrieval and processes large mail archives (700+ MB), implying Python dependencies for that specific server.
Verified SafeView Analysis
The provided source code primarily consists of documentation and Antora configuration files. It does not contain executable code, 'eval' statements, obfuscation, or hardcoded secrets directly. The project describes interacting with external services (GitHub, Atlassian, mail archives) via underlying MCP servers. The security of these actual MCP server implementations (e.g., mail-mcp), particularly concerning credential handling and network interactions, is not assessable from the truncated code and would require separate audits.
Updated: 2026-01-11GitHub→
0
0
Low Cost
haakco icon

mcp-utils

by haakco

Sec7

Provides a comprehensive set of shared utilities for building robust and standardized Model Context Protocol (MCP) servers, covering architecture, formatting, data operations, caching, rate limiting, WebSockets, and logging.

Setup Requirements

  • ⚠️Requires Node.js v18.0.0 or higher.
  • ⚠️Local development with other MCP servers may require configuring `file:` dependencies in `package.json`.
  • ⚠️Logging utilities (`createLoggerFromEnv`) rely on `DEBUG` and `LOG_LEVEL` environment variables for configuration.
Verified SafeView Analysis
The `postman-helpers.ts` file includes a `validateScript` function that uses `new Function(script)` for basic JavaScript syntax validation. While not direct execution, compiling arbitrary, untrusted input with `new Function()` can be a security risk due to potential denial-of-service (resource exhaustion during compilation) or obscure V8 engine exploits. It is generally safer to use dedicated JavaScript parser libraries for syntax validation without runtime compilation. No other critical vulnerabilities like hardcoded secrets or malicious patterns were identified in the provided code.
Updated: 2026-01-07GitHub→
0
0
Medium Cost
Rishabh899 icon
Sec9

This server acts as an expense tracker, allowing users to add, list, and summarize financial expenditures using a FastMCP API.

Setup Requirements

  • ⚠️Requires Python 3.11 or higher, as specified in `pyproject.toml`.
  • ⚠️The SQLite database (`expenses.db`) is created in a temporary directory, meaning all expense data will be lost upon system reboot or cleanup of temporary files. This design makes the data non-persistent across sessions.
  • ⚠️The `categories.json` file is expected in the same directory as `main.py` for custom categories. While defaults are provided if not found, custom category management requires this file placement.
Verified SafeView Analysis
The code employs parameterized queries for all database interactions, effectively preventing SQL injection vulnerabilities. There are no instances of `eval` or code obfuscation. No hardcoded secrets or API keys are present. The database is initialized and managed in a temporary directory, which is generally safe for isolation, though it means data is non-persistent. The server binds to `0.0.0.0`, requiring external network security if exposed publicly.
Updated: 2025-11-20GitHub→
0
0
Low Cost
remysaissy icon

mcp-unix-tools

by remysaissy

Sec7

A local Model Context Protocol (MCP) server that exposes a collection of Unix tools to AI models for local development and automation tasks.

Setup Requirements

  • ⚠️Rust toolchain (2024 edition or later) and Cargo required
  • ⚠️Unix-like environment (Linux, macOS, or WSL on Windows) is mandatory
  • ⚠️Provided source for `src/main.rs` is currently a 'Hello world.'; actual Unix tool implementations are not yet present in the snippet.
Verified SafeView Analysis
The project is explicitly designed for local development and warns against exposing the server to untrusted networks. It states principles of 'Input Validation' and 'Sandboxing' for future Unix tool implementations. However, the provided `src/main.rs` is currently a 'Hello world.', so the actual implementation of command execution and its security hardening measures are not visible for direct audit. Running arbitrary commands via an AI still presents inherent risks if not perfectly isolated and validated.
Updated: 2025-11-20GitHub→
0
0
Medium Cost
NikhilAdvani icon
Sec9

Manages personal expenses by allowing users to add, list, summarize, edit, delete, and search expense entries.

Setup Requirements

  • ⚠️Database stored in a temporary directory (`/tmp` or similar) and will not persist across system reboots or temporary directory cleanups, leading to potential data loss.
  • ⚠️Requires Python 3.12+ due to project configuration.
Verified SafeView Analysis
SQL queries use parameterized statements, effectively preventing SQL injection vulnerabilities. No 'eval' or other dangerous code patterns are present. No hardcoded secrets or sensitive API keys were found. The database is stored in a temporary directory, which while generally writable, could be a minor concern regarding persistence and isolation in multi-user environments.
Updated: 2025-12-13GitHub→
0
0
Medium Cost
Bazilio-san icon

mcp-atlassian-ts

by Bazilio-san

Sec9

Automate and extend Atlassian JIRA and Confluence operations for AI agents using the Model Context Protocol (MCP).

Setup Requirements

  • ⚠️Requires valid Atlassian JIRA and/or Confluence URL and credentials (username/password or Personal Access Token) in the .env file.
  • ⚠️Mandatory `MCP_SERVICE_MODE` environment variable must be set to 'jira' or 'confluence' for the server to start.
  • ⚠️Default port 3000 might be in use; requires changing `SERVER_PORT` environment variable if conflict occurs.
Verified SafeView Analysis
The server employs multiple authentication methods (Basic, PAT, OAuth2), uses `helmet` for security headers, and `rate-limiter-flexible` to prevent abuse. Sensitive configurations are correctly externalized to `.env` files. The user substitution (impersonation) feature, while powerful, requires careful administrator setup to prevent unauthorized access. CORS is limited to development environments, which is appropriate. No `eval` or obfuscation detected. Strong error handling and logging contribute to overall security and maintainability.
Updated: 2025-11-26GitHub→
0
0
Medium Cost
yohanesmario icon

atlassian-mcp

by yohanesmario

Sec8

Automates interactions with Atlassian Jira and Confluence by providing a custom, unified MCP server with optimized token usage for tool definitions and robust Markdown to Atlassian Document Format (ADF) conversion.

Setup Requirements

  • ⚠️Requires ATLASSIAN_EMAIL, ATLASSIAN_API_TOKEN, and ATLASSIAN_DOMAIN environment variables to be set.
  • ⚠️If using a .env file for credentials, it must have secure permissions (e.g., `chmod 600 .env`) or it will be silently skipped.
  • ⚠️The ATLASSIAN_DOMAIN must be in the format `company.atlassian.net` (domain only, no protocol or path).
Verified SafeView Analysis
The server demonstrates good security practices including input validation (e.g., issue keys, page IDs), enforcement of TLS 1.2 for network requests, and a critical check for secure permissions (chmod 600) on the .env file. API tokens are passed via standard Basic Auth, relying on environment variable security. Potential risks include the inherent complexity of markdown-to-ADF parsing which could theoretically be vulnerable to Regular Expression Denial of Service (ReDoS) if exposed to extremely large and malicious inputs, though Go's regex engine is generally robust. Attachment uploads are handled carefully with filename sanitization and size limits. No 'eval' or direct code execution from user input is apparent.
Updated: 2025-12-15GitHub→
PreviousPage 672 of 760Next