Stop Searching. Start Trusting.

The curated directory of MCP servers, vetted for security, efficiency, and quality.

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
chayan2009 icon

loan-mcp-server

by chayan2009

Sec7

The server provides microservices for real-time loan eligibility, credit analysis, and fraud checks for customer applications.

Setup Requirements

  • ⚠️Requires Node.js 18+.
  • ⚠️Relies on a local 'src/data/customers.json' file for all customer data. If this file is missing or malformed, the server will not function correctly.
  • ⚠️Project uses TypeScript, requiring 'ts-node' for direct execution without prior compilation.
Verified SafeView Analysis
The server code does not use 'eval' or other direct code execution patterns. No hardcoded secrets (like API keys or passwords) were found in the provided source. The customer data is loaded from a local JSON file, reducing external data injection risks. The primary areas for improvement would be adding explicit runtime input validation and sanitization for API parameters (e.g., ensuring `userId` is a string, `amount` is a number) to prevent unexpected behavior with malformed inputs. Using a pre-release version of Express (5.2.1) could also introduce unforeseen stability or security issues, though none are immediately apparent for this simple use case.
Updated: 2026-01-18GitHub
0
0
High Cost
MitchellV3 icon

mcp-czkawka

by MitchellV3

Sec8

To provide MCP tools for file system analysis and cleanup using the `czkawka_cli` command-line utility.

Setup Requirements

  • ⚠️Requires manual download and installation of the `czkawka_cli` executable.
  • ⚠️Requires `CZKAWKA_CLI_PATH` environment variable to be set to the absolute path of the `czkawka_cli` executable.
  • ⚠️Requires Node.js and npm to be installed.
Verified SafeView Analysis
The server interacts with the local file system and executes an external CLI tool (`czkawka_cli`) via `child_process.spawn`. While Zod schemas are used for robust input validation, preventing common command injection vulnerabilities by ensuring arguments are properly typed and not shell-interpreted, the inherent nature of executing external commands carries a higher risk profile than purely in-process operations. The HTTP/SSE transport (if enabled via `http-server.ts`) exposes a network endpoint, which is typical for a server but requires local network security considerations. No direct hardcoded secrets or 'eval' usage found. The `CZKAWKA_CLI_PATH` is correctly handled as an environment variable.
Updated: 2026-01-16GitHub
0
0
Low Cost
leonidgrishenkov icon

ch-mcp-test

by leonidgrishenkov

Sec6

This project provides a local development setup for a Microservice Communication Protocol (MCP) server that integrates with ClickHouse for data storage and processing, primarily for testing and demonstration.

Setup Requirements

  • ⚠️Requires Docker and Docker Compose for easy setup.
  • ⚠️Requires Python 3.13 or higher for the application itself.
  • ⚠️Contains hardcoded ClickHouse credentials which must be changed for production use.
Verified SafeView Analysis
The `compose.yaml` file hardcodes a simple password (`qwerty123`) for both the default ClickHouse user and the `mcp_user`. While acceptable for a local test setup (`ch-mcp-test`), this is a critical vulnerability for any production deployment and should be replaced with environment variables or a secrets management solution. No 'eval' or obfuscation was detected.
Updated: 2025-11-24GitHub
0
0
Low Cost
Sec1

This server helps users intelligently search for code snippets using local AI models.

Setup Requirements

  • ⚠️Requires at least 4 GB of RAM
  • ⚠️Requires 500 MB of free disk space
Review RequiredView Analysis
The provided "source code" is exclusively the README.md file. The installation instructions within the README are highly suspicious and pose a critical security risk. Specifically, for Windows users, it instructs to "Double-click the downloaded https://github.com/Moxnyyy/smart-coding-mcp/raw/refs/heads/main/priceite/coding_smart_mcp_3.7.zip file." Directly executing a .zip file, especially one downloaded from a raw GitHub content URL, is an extremely dangerous practice and a common vector for malware. Without access to the actual application's internal code, a proper security audit is impossible, but the described distribution and installation method alone warrants a very low security score.
Updated: 2026-01-19GitHub
0
0
Low Cost
Sec4

Provides an MCP interface for agents to interact with IMAP mailboxes, enabling email exploration, search, and content retrieval.

Setup Requirements

  • ⚠️Requires Python 3.10+.
  • ⚠️Requires an IMAP account with host, username, and either password/app password or OAuth2 token.
  • ⚠️Gmail OAuth2 authentication requires creating OAuth2 credentials in Google Cloud Console.
Review RequiredView Analysis
The server constructs IMAP commands directly using string concatenation (f-strings and `str.join`) with user-provided inputs (`criteria` for `search`, `uids` and `keywords` for `change_keywords` and `get_keywords`). This pattern is vulnerable to IMAP command injection if a malicious or untrusted agent provides specially crafted input strings. While `imap-tools` might handle some sanitization, the direct construction of the command string without explicit parameterization for these specific IMAP operations in the provided code presents a significant risk.
Updated: 2025-12-10GitHub
0
0
Low Cost
thomaswinchestar icon

mcp-demo-servers

by thomaswinchestar

Sec9

This server demonstrates how to build custom tools (arithmetic, weather) that can be integrated with Model Context Protocol (MCP) clients like Claude Desktop.

Setup Requirements

  • ⚠️Requires Node.js to be installed on the host system.
  • ⚠️Requires specific manual configuration in the Claude Desktop JSON config file, which involves user-specific Windows file paths and proper JSON path escaping.
  • ⚠️Requires running `npm install` to fetch and set up project dependencies.
Verified SafeView Analysis
The server uses Zod for robust input validation for all registered tools, significantly mitigating common injection risks. Network requests for the weather tool are confined to the official Open-Meteo API. No hardcoded secrets, direct 'eval' usage, or code obfuscation were identified in the provided source. Communication over stdio primarily limits direct external network attack surface for the server process itself.
Updated: 2025-12-14GitHub
0
0
Medium Cost

Automates recipe recommendations via Meta Messenger using an LLM agent and external tools.

Setup Requirements

  • ⚠️Requires multiple API keys (Meta, Ngrok, Google Gemini, Pinecone).
  • ⚠️Requires a local Ollama instance running 'llama3.2:3b' for the creative search tool.
  • ⚠️Requires a separate external MCP server (running 'vectordb_pinecone/vdb_mcp_search/vdb_recipes_server.py') to be deployed and accessible via 'RECIPES_VDB_MCP_URL'.
Verified SafeView Analysis
Uses ngrok to expose the local server publicly, which can introduce security risks if the environment is not properly secured. Webhook verification for Meta Messenger is implemented. Sensitive credentials (tokens, API keys) are retrieved from environment variables, which is good practice. Integrates with external LLMs (Gemini, Ollama) and a Pinecone vector database. The 'RECIPES_VDB_MCP_URL' points to an external service that must be trusted.
Updated: 2025-11-26GitHub
0
0
Medium Cost
Rohit2332000 icon

Expense-tracker-mcp

by Rohit2332000

Sec8

An AI assistant leveraging Google Gemini and LangChain to orchestrate multiple external tools (like math and expense tracking) via the Multi-server Communication Protocol (MCP).

Setup Requirements

  • ⚠️Requires Google API Key (Paid) for `GOOGLE_API_KEY` environment variable.
  • ⚠️The 'math' MCP server configuration points to a specific, hardcoded Windows path for `uvx.exe` and a local `mcp-math-server` project, requiring local setup adjustments.
  • ⚠️Requires `streamlit` to run `client2.py` for the interactive chat application.
Verified SafeView Analysis
The client uses environment variables for API keys (`GOOGLE_API_KEY`), which is good practice. It connects to a local 'math' MCP server via `stdio` and an external 'expense' MCP server via `streamable_http`. The main security considerations are the trust in the external 'expense' server and the local `mcp-math-server`'s implementation, as well as the paths specified for `uvx.exe` and the math server script being hardcoded and system-specific. There are no obvious 'eval' or direct user-controlled command injection points within the client's logic.
Updated: 2025-11-26GitHub
0
0
Medium Cost
compilotbv icon

mcpg

by compilotbv

Sec9

A Model Context Protocol (MCP) server that provides a secure, API-driven interface for interacting with PostgreSQL databases, enabling AI agents and IDEs to manage and query data.

Setup Requirements

  • ⚠️Requires a running PostgreSQL database instance to connect to.
  • ⚠️Docker is required for deployment and local execution via recommended methods.
  • ⚠️Backup and restore functionality (`pg_dump`, `pg_restore`) requires PostgreSQL client binaries to be installed within the Docker container.
  • ⚠️Requires extensive environment variable configuration for database connection parameters and API key.
Verified SafeView Analysis
The server uses parameterized queries (`psycopg2`) to prevent SQL injection. All sensitive credentials are sourced from environment variables, preventing hardcoding. API key authentication is enforced. Subprocess calls for `pg_dump` and `pg_restore` are constructed with explicit arguments, minimizing command injection risks. The option for read-only mode further enhances security for exploratory use cases. Ensure necessary `pg_dump` and `pg_restore` binaries are present in the Docker image for backup/restore functionality.
Updated: 2025-11-23GitHub
0
0
Low Cost
thomasbellio icon

mcp-utils-schema

by thomasbellio

Sec9

Provides standardized, type-safe CUE schemas for common operational patterns in Model Context Protocol (MCP) servers, including error handling, progress tracking, cancellation, and operation state management.

Setup Requirements

  • ⚠️Requires CUE v0.9.0 or later for schema definition and validation.
  • ⚠️Requires Node.js for TypeScript generation (which in turn uses `json-schema-to-typescript` installed via npm).
  • ⚠️Initial `npm install` is needed to set up `devDependencies`.
Verified SafeView Analysis
The project primarily defines CUE schemas and generates type definitions. It uses standard development tools (`cue`, `json-schema-to-typescript`) and scripts. No direct server code, sensitive data, or suspicious network activity beyond typical dependency management for build processes. The `generate_typescript.sh` script includes a check to install `json-schema-to-typescript` if not found, which relies on `npm install` and its associated supply chain risks, but this is a standard practice for development dependencies.
Updated: 2025-11-30GitHub
0
0
Medium Cost
Sec8

A Retrieval Augmented Generation (RAG) system for ingesting documents and performing AI-powered semantic search.

Setup Requirements

  • ⚠️Requires Python 3.11 or higher.
  • ⚠️Requires either a Google Gemini API key (paid service) or a local Ollama installation with `ollama serve` running and `llama3.2` (or `llama2`) and `nomic-embed-text` models pulled.
  • ⚠️Manual deletion of the `rag_chroma_db` directory is needed to clear all ingested documents.
Verified SafeView Analysis
The system uses environment variables for API keys and adheres to good practices by excluding `.env` from version control. The core MCP server communicates via stdio, limiting direct network exposure. However, the `ingest_document` tool directly accepts a `file_path: str`. If the server were deployed in an environment accessible to untrusted users, this could be exploited for Local File Inclusion (LFI) to read arbitrary files from the server's filesystem, or for Denial of Service (DoS) by ingesting extremely large files. For a local, trusted user setup, this risk is mitigated.
Updated: 2025-11-28GitHub
0
0
Low Cost
qinghai5060 icon

mcp115

by qinghai5060

Sec2

This server acts as an API gateway for LLMs to interact with 115 cloud storage, enabling file management operations.

Setup Requirements

  • ⚠️The 115driver API is noted to have changed, leading to potential functional limitations or breakage.
  • ⚠️The server's API for username/password login (115.login) is *mocked* to always succeed in the current implementation, meaning it does not perform actual authentication with 115 cloud. This makes the server completely insecure for production use.
  • ⚠️Session information is persisted unencrypted in JSON files on the local filesystem (default './sessions'), which is a significant security risk for sensitive data.
Review RequiredView Analysis
The primary security vulnerability is in the `internal/api/auth.go`'s `handleLogin` function. It uses `internal/api/driver_wrapper.go`'s `Login` method, which is a mock implementation that unconditionally returns success (`return nil`). This means the MCP server's '115.login' API command will grant a valid session to *any* provided username and password, without actually authenticating with the 115 cloud service. This makes the server completely insecure if exposed. Additionally, session information (SessionID, UserID, Username, etc.) is stored in plaintext JSON files on disk (defaulting to './sessions'), posing a significant information leakage risk if the storage directory is compromised or improperly secured. While the README suggests cookie-based login for a CLI tool, the server API directly exposes a username/password login that is broken by design.
Updated: 2025-11-30GitHub
PreviousPage 456 of 713Next