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)

30
2
Medium Cost
GerritCodeReview icon

gerrit-mcp-server

by GerritCodeReview

Sec8

This server allows a language model (like Gemini) to query, retrieve details, and manage code reviews within the Gerrit system.

Setup Requirements

  • ⚠️Requires Python 3.12+ (or 3.11+ as per README, but pyproject.toml specifies >=3.12).
  • ⚠️Requires 'curl' to be installed and available in PATH.
  • ⚠️Manual creation and configuration of `gerrit_mcp_server/gerrit_config.json` with Gerrit credentials is required.
  • ⚠️The build process uses 'uv', which is installed during `./build-gerrit.sh` but implicitly required for that script to function correctly in some scenarios (e.g., if `.venv` is missing and needs rebuilding).
Verified SafeView Analysis
The server primarily interacts with Gerrit via `curl` commands. The `run_curl` function uses `asyncio.create_subprocess_exec` with a list of arguments, which mitigates shell injection risks. Input parameters for Gerrit queries are URL-quoted. Authentication details (username, auth_token, gitcookies_path) are stored in `gerrit_config.json`, which must be secured by the user. The `server.log` could expose executed commands and responses, requiring local filesystem security. Overall, the approach to command execution is robust against common injection vectors, but secure handling of the configuration file is critical.
Updated: 2026-01-08GitHub→
30
1
Medium Cost
sgaunet icon

gitlab-mcp

by sgaunet

Sec9

Provides GitLab integration tools as a Model Context Protocol (MCP) server for Claude Code, enabling AI to manage issues, labels, epics, and project metadata.

Setup Requirements

  • ⚠️Requires `GITLAB_TOKEN` environment variable with appropriate GitLab API scopes (`api`, `read_api`, `write_api`).
  • ⚠️Requires Go 1.21 or later if building from source or running directly via `go run .`.
  • ⚠️Designed to be integrated with Claude Code CLI as an MCP server, not typically run as a standalone network service.
Verified SafeView Analysis
The server uses `GITLAB_TOKEN` from environment variables, which is good practice. Communication occurs over standard input/output (stdio), significantly reducing external network attack surfaces. No `eval` or obvious malicious patterns were found. Label validation can be disabled via `GITLAB_VALIDATE_LABELS=false`, which is a documented user-configurable option and not a vulnerability.
Updated: 2026-01-10GitHub→
30
1
High Cost
mcp-z icon

mcp-pdf

by mcp-z

Sec7

Generate PDFs from text, layouts, or JSON Resume data, render PDF pages as images, and measure text dimensions for layout planning.

Setup Requirements

  • ⚠️Requires Node.js version 20 or higher.
  • ⚠️Emoji rendering requires `NotoColorEmoji.ttf` to be available; it's typically installed automatically via `postinstall` script, but manual intervention might be needed if postinstall fails.
  • ⚠️`@napi-rs/canvas` is a native module that requires pre-built binaries for your specific OS/architecture or successful compilation during installation.
Verified SafeView Analysis
The server uses `eval` for dual ESM/CJS module loading, which is generally discouraged but used here in a controlled context for internal module initialization. `LiquidJS` is used for templating in `pdf-resume`; while generally sandboxed, custom templates could pose a risk if not carefully reviewed. File system operations for temporary files and resource storage are present, but inputs are validated by `zod` and `ajv` schemas, and network-downloaded fonts are placed in temporary directories. No hardcoded secrets or direct remote code execution vulnerabilities are apparent.
Updated: 2026-01-16GitHub→
30
1
Low Cost
Jing-yilin icon
Sec9

Provides structured, token-efficient access to Product Hunt data via its GraphQL API for AI models and agents.

Setup Requirements

  • ⚠️Requires a Product Hunt API access token (`PRODUCTHUNT_ACCESS_TOKEN` or `PH_ACCESS_TOKEN`) which needs to be generated from the Product Hunt API Dashboard.
  • ⚠️Requires Node.js runtime (version >= 18.0.0).
Verified SafeView Analysis
The server correctly handles API keys via environment variables and makes standard HTTPS requests. It includes a feature to save cleaned data to a local directory specified by the `save_dir` argument. While this is a local operation, it's generally safe within the intended MCP server context where the LLM controls the arguments. No 'eval' or direct arbitrary code execution patterns were found.
Updated: 2026-01-10GitHub→
30
1
Medium Cost
hyperpolymath icon

poly-ssg-mcp

by hyperpolymath

Sec3

Unified server for managing and interacting with various static site generators written in diverse programming languages, designed for integration with AI agents.

Setup Requirements

  • ⚠️Requires Deno runtime environment.
  • ⚠️Each supported Static Site Generator (SSG) binary (e.g., `zola`, `stack`, `julia`, `sbcl`, `mix`) must be individually installed and available in the system's PATH for its adapter to connect and function.
  • ⚠️Requires `rescript` to be installed and accessible via Deno's npm compatibility layer for local development and build processes (`deno run -A npm:rescript build`).
Review RequiredView Analysis
The server uses `Deno.Command.run` to execute external SSG binaries. Critical command injection vulnerabilities exist in adapters like `Franklin`, `Documenter`, `StaticWebPages`, and `Coleslaw` where user-provided input (`path`, `template`, `title`) is directly interpolated into shell commands for `julia -e` or `sbcl --eval`. While `Executor.res.js` contains a `sanitizeArg` function, it is not consistently applied in all adapter execution functions, leaving multiple paths for arbitrary command execution. This is a severe risk.
Updated: 2026-01-18GitHub→
30
1
Medium Cost
Doist icon

twist-ai

by Doist

Sec9

Enables AI agents to access and interact with a Twist workspace, providing tools for managing threads, conversations, and user information.

Setup Requirements

  • ⚠️Requires `TWIST_API_KEY` environment variable to be set.
  • ⚠️`@modelcontextprotocol/sdk` is a peer dependency and must be explicitly installed by the user.
  • ⚠️Requires Node.js 18 or higher.
Verified SafeView Analysis
The server uses `process.env.TWIST_API_KEY` for API authentication, avoiding hardcoded secrets. Input validation is performed using Zod schemas. It relies on the official `@doist/twist-sdk` and `@modelcontextprotocol/sdk`. No 'eval' or obvious obfuscation found. Standard API interactions mean data is transmitted over the network, which is expected for its functionality.
Updated: 2026-01-19GitHub→
30
1
Medium Cost
neverinfamous icon

db-mcp

by neverinfamous

Sec9

An enterprise-grade Model Context Protocol (MCP) server providing advanced database operations, AI-native functionalities like vector search, and OAuth 2.1 authentication for SQLite databases.

Setup Requirements

  • ⚠️Requires Node.js 24+ (LTS).
  • ⚠️Docker is recommended for deployment and required for the `--sqlite-native` backend in certain IDEs or environments.
  • ⚠️Enabling OAuth 2.1 requires a separate Authorization Server (e.g., Keycloak) to be set up and configured, involving several manual steps.
  • ⚠️The `--sqlite-native` backend, offering more tools including transactions and window functions, relies on the `better-sqlite3` optional dependency, which may require specific build environments or manual installation.
Verified SafeView Analysis
The project demonstrates a strong commitment to security, including explicit SQL injection prevention via parameter binding, OAuth 2.1 compliance (RFC 9728/8414), strict TypeScript usage to prevent common bugs, and secure handling of sensitive configurations (e.g., environment variables for secrets, redaction in logs). CORS defaults to restrictive `origin: false`. CodeQL static analysis is integrated. No 'eval' or malicious patterns were found in the provided source code.
Updated: 2026-01-19GitHub→
30
1
Low Cost
Imashanilupul icon
Sec8

Provides open-source intelligence (OSINT) reconnaissance capabilities as a Model Context Protocol (MCP) server for integration with AI agents.

Setup Requirements

  • ⚠️Requires Python 3.12 or higher (as per pyproject.toml).
  • ⚠️Requires a Wigle API account and API keys (WIGLE_API_NAME, WIGLE_API_TOKEN) for BSSID lookups.
  • ⚠️Requires creating a `.env` file for API keys in the project root.
Verified SafeView Analysis
The server's core functionality involves direct network connections and external API calls, which are inherent to OSINT tools. API keys are handled securely via environment variables using pydantic-settings and dotenv. Broad `except` blocks in tool functions could mask errors, but no 'eval' or other severe vulnerabilities like code injection were found. It is safe to run when used responsibly and ethically, understanding its purpose as a reconnaissance tool.
Updated: 2025-11-24GitHub→
30
1
Medium Cost
gesslar icon

fluffos-mcp

by gesslar

Sec8

An MCP server that enables AI assistants to perform driver-level validation, debugging, and documentation lookup for LPC code using FluffOS CLI tools.

Setup Requirements

  • ⚠️Requires FluffOS installed with `symbol` and `lpcc` binaries available.
  • ⚠️Requires Node.js v22 or higher.
  • ⚠️Requires environment variables FLUFFOS_BIN_DIR, MUD_RUNTIME_CONFIG_FILE, and optionally FLUFFOS_DOCS_DIR to be set with absolute paths.
Verified SafeView Analysis
The server leverages Node.js `child_process.spawn` to execute external FluffOS CLI tools (`symbol`, `lpcc`, and a `search_docs.sh` script). Arguments to these tools, including file paths and search queries, are passed from AI input. While the Node.js code itself uses array arguments for `spawn` (mitigating shell injection in the initial call) and the shell script explicitly double-quotes its `$QUERY` argument for `grep`/`rg`, the overall security relies heavily on the trustworthiness of the configured FluffOS binaries and documentation directory, as well as the robustness of the underlying FluffOS CLI tools when processing potentially untrusted or malformed file paths.
Updated: 2026-01-18GitHub→
30
17
Medium Cost
mcp-use icon
Sec4

This server provides interactive tools and React widgets for exploring and querying your Supabase database and viewing its service status.

Setup Requirements

  • ⚠️Requires Node.js 18+ and yarn.
  • ⚠️Requires an active Supabase project.
  • ⚠️Requires a Supabase Personal Access Token (which can incur costs depending on usage).
Review RequiredView Analysis
The server uses `process.env.ACCESS_TOKEN` for Supabase authentication, which is good practice. However, the `SupabaseStatusWidget` uses `dangerouslySetInnerHTML` to render descriptions from an external RSS feed (`https://status.supabase.com/history.rss`). This is a critical Cross-Site Scripting (XSS) vulnerability if the RSS feed content can be manipulated by an attacker, potentially allowing arbitrary script execution within the widget. While `CSP_URLS` is configured for widget loading, `dangerouslySetInnerHTML` needs explicit sanitization of the content itself to prevent injection risks.
Updated: 2025-12-18GitHub→
30
1
High Cost
ltsch icon
Sec10

A specialized MCP server for Cursor IDE that automatically creates and manages project-specific rule directories with automatic file watching and AI-powered rule management tools.

Setup Requirements

  • ⚠️Requires Node.js 18+ (automated by 'install.sh' script).
  • ⚠️Requires Cursor IDE to be installed and configured for MCP.
  • ⚠️When using 'generate-config.js' for new projects, paths in the generated '.cursor/mcp.json' output may require manual adjustment to match the actual server repository location.
Verified SafeView Analysis
The server is explicitly designed for local-first operation with zero network connectivity. It communicates solely via stdio (standard input/output) with Cursor IDE, eliminating network attack surfaces. It does not use 'eval', and there are no apparent hardcoded secrets. AI-powered features leverage Cursor IDE's existing LLM integration rather than initiating external network calls directly from the server process. File system operations are confined to project-specific '.cursor/rules' directories.
Updated: 2025-11-28GitHub→
30
1
Medium Cost
findmine icon

findmine-mcp

by findmine

Sec9

An MCP server that provides product styling, outfitting recommendations, visually similar product suggestions, and fashion advice by integrating with the FindMine API for LLM-powered shopping stylists.

Setup Requirements

  • ⚠️Requires a FindMine Application ID (`FINDMINE_APP_ID`) for non-demo usage, which must be obtained from FindMine.
  • ⚠️User interaction tracking (`track_interaction` tool) and item detail updates (`update_item_details` tool) are disabled by default and require explicit environment variables (`FINDMINE_ENABLE_TRACKING=true`, `FINDMINE_ENABLE_ITEM_UPDATES=true`) to be enabled.
  • ⚠️Requires an MCP-compatible client (e.g., Claude Desktop) to connect and interact with the server after installation.
Verified SafeView Analysis
The server demonstrates good security practices including input validation via Zod schemas for all tool arguments, use of environment variables for API keys and configurations (with safe defaults), and logging to stderr to avoid interfering with the MCP protocol. The API client handles network requests with retries and basic error handling, constructing URLs and JSON bodies appropriately to prevent common injection vectors. Potentially sensitive features like tracking and item updates are opt-in via environment variables. No 'eval' or malicious obfuscation patterns were found. The primary risk would be in the security of the upstream FindMine API or misconfiguration of environment variables.
Updated: 2025-12-03GitHub→
PreviousPage 220 of 713Next