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
hdmt icon
Sec9

Manages HubSpot marketing emails (list, get, create draft, update) from Claude Desktop.

Setup Requirements

  • ⚠️Requires obtaining a HubSpot Access Token with 'content' scope, involving several manual steps in the HubSpot portal.
  • ⚠️Requires manual editing of the Claude Desktop configuration file (~/Library/Application Support/Claude/claude_desktop_config.json).
  • ⚠️The 'command' in Claude Desktop config requires the absolute path to the 'hubspot-email-mcp' directory.
Verified SafeView Analysis
The server correctly retrieves the HubSpot Access Token from an environment variable (HUBSPOT_ACCESS_TOKEN), avoiding hardcoded secrets. Input validation is performed using Zod schemas. It explicitly states that email sending functionality is not implemented to prevent accidental sends, which is a significant security consideration. Communication with the HubSpot API uses HTTPS. No 'eval' or other highly dangerous patterns are observed. The primary security responsibility lies with the user in securely managing their HubSpot Access Token.
Updated: 2025-11-30GitHub→
0
0
Medium Cost
Sec1

The Snowflake MCP Server provides a comprehensive suite of tools for managing, monitoring, and optimizing Snowflake data clouds, encompassing data discovery, intelligence, automation, security, and governance.

Setup Requirements

  • ⚠️Requires Docker and Docker Compose for deployment.
  • ⚠️Requires Snowflake ACCOUNTADMIN role credentials (account, user, base64-encoded DER private key).
  • ⚠️Requires specific firewall port (8083 by default) to be open for access.
Review RequiredView Analysis
The server is configured to run with `ACCOUNTADMIN` privileges, exposing full control over the Snowflake environment via its API. The `run_snowflake_query` tool (and potentially others) executes arbitrary SQL strings directly without parameter binding, making it highly vulnerable to SQL injection. Furthermore, the `server_http.py`'s CORS policy `allow_origins=["*"]` enables access from any domain, significantly broadening the attack surface. Snowflake private keys are stored as environment variables, which, while more secure than passwords, are still sensitive credentials.
Updated: 2025-11-27GitHub→
0
0
Low Cost
nitvob icon
Sec9

Provides real-time weather forecasts and alerts for US locations by integrating with the National Weather Service API.

Setup Requirements

  • ⚠️Requires Python 3.13 or higher.
  • ⚠️Requires the `uv` package manager for installation and execution.
  • ⚠️Requires internet access to communicate with the National Weather Service API.
  • ⚠️Only provides data for US locations.
Verified SafeView Analysis
The server uses `httpx` for network requests to a public API (api.weather.gov) which requires no API key. There are no hardcoded secrets, `eval` calls, or other obvious vulnerabilities. Error handling for network requests is broad but returns `None` which is handled by the calling functions, preventing direct exposure of internal errors.
Updated: 2026-01-19GitHub→
0
0
Low Cost
andreasgerstmayr icon

tempo-mcp-gateway

by andreasgerstmayr

Sec9

The tempo-mcp-gateway acts as an instance-aware and tenant-aware gateway for Grafana Tempo's MCP (Management Control Plane) server, simplifying interaction with multiple Tempo instances within a Kubernetes cluster.

Setup Requirements

  • ⚠️Requires a Kubernetes cluster for deployment and operation.
  • ⚠️Relies on Grafana Tempo Operator Custom Resource Definitions (CRDs) (TempoStack, TempoMonolithic) to discover Tempo instances.
  • ⚠️The provided deployment manifests include OpenShift-specific resources (Route), implying an OpenShift environment, though it can be adapted for standard Kubernetes.
Verified SafeView Analysis
The server uses standard Kubernetes practices for secure communication, including in-cluster configuration for the Kubernetes API client and a service CA certificate for TLS. It properly forwards client authorization tokens to downstream Tempo instances, maintaining the client's original permissions. There are no observed hardcoded secrets, 'eval' usage, or other obvious malicious patterns. The RBAC rules are appropriately scoped to 'list' Tempo custom resources, limiting its own cluster permissions. The filtering of accessible tenants by probing readiness endpoints is a good practice to ensure the gateway only exposes what the client is authorized to access.
Updated: 2025-12-15GitHub→
0
0
High Cost
slop-at icon

slopnet-mcp

by slop-at

Sec7

This server allows users to post markdown notes ('slops') to a distributed knowledge graph, automatically extracting entities and integrating with a user's GitHub repository.

Setup Requirements

  • ⚠️Requires `uv` (a modern Python package installer) to be installed locally.
  • ⚠️Requires users to create and configure a GitHub repository for storing their slops.
  • ⚠️Requires the Claude Desktop application and manual modification of its `claude_desktop_config.json`.
  • ⚠️The GLiNER2 entity extraction model (340M parameters) will be downloaded on first use, requiring significant disk space and an internet connection.
Verified SafeView Analysis
The server uses `subprocess.run` for Git operations (clone, add, commit, push), which are necessary for its functionality and appear to be well-controlled with predefined commands and arguments. It makes network requests to 'https://slop.at' (default) for knowledge graph queries, updates, and publishing slop content and extracted entities. This means user data (slops, extracted entities, RDF triples) is sent to an external service, which is inherent to its distributed design. SPARQL queries and updates are passed directly to the remote server. While `query_slops` is for data retrieval, `update_graph` explicitly warns 'Advanced users only - use with caution!', as arbitrary SPARQL UPDATE queries could be destructive if the backend server is not properly secured against such operations or if a malicious prompt were to generate one. No `eval` or obvious hardcoded secrets were found.
Updated: 2026-01-19GitHub→
0
0
High Cost
7elu-dl icon
Sec9

This server connects Claude AI to an OpenCTI threat intelligence platform, allowing the AI to query, analyze, and manage cybersecurity intelligence data.

Setup Requirements

  • ⚠️Requires an existing OpenCTI instance (v5.x or v6.x) to be deployed and accessible.
  • ⚠️An OpenCTI API token with appropriate permissions is mandatory.
  • ⚠️Requires Python 3.10 or higher.
Verified SafeView Analysis
The project demonstrates a strong focus on security best practices: API tokens are loaded from environment variables (`.env`) and explicitly warned against committing to version control. SSL verification is enabled by default for all HTTP communications. Input validation is performed for observable types and GraphQL queries use variables, reducing direct injection risks. There are no signs of 'eval', obfuscation, or arbitrary code execution. The documentation provides clear security notes on token management and SSL usage.
Updated: 2025-11-26GitHub→
0
0
Medium Cost
RaphaEnterprises-AI icon

argus-mcp-server

by RaphaEnterprises-AI

Sec9

Model Context Protocol (MCP) server for Argus E2E Testing Agent, enabling AI coding assistants to interact with web automation, E2E testing, and quality intelligence capabilities.

Setup Requirements

  • ⚠️Requires a Cloudflare account for deployment and utilization of Workers, R2, KV, and Durable Objects.
  • ⚠️Dependencies on external Argus API (ARGUS_API_URL) and Argus Brain API (ARGUS_BRAIN_URL) for core functionality (browser automation, AI intelligence, test generation, etc.). These services are not part of this repository.
  • ⚠️Requires configuration of several Cloudflare bindings (Workers AI, R2 bucket, KV namespace, Durable Objects) and secrets (API_TOKEN, BROWSER_POOL_JWT_SECRET, ANTHROPIC_API_KEY) for full operation.
Verified SafeView Analysis
The server uses JWT for authenticating with its browser pool, signed URLs for R2 screenshot access, and an OAuth2 device flow for user authentication, which are good security practices. Secrets like API tokens and JWT signing keys are stored as environment variables/Cloudflare secrets. A minor concern is the 'default-secret-for-dev' fallback for BROWSER_POOL_JWT_SECRET, which could be a risk if inadvertently used in production, though it's clearly labeled for development.
Updated: 2026-01-17GitHub→
0
0
Medium Cost
bioanywhere icon

happy

by bioanywhere

Sec8

This repository provides client libraries for interacting with the World News API, enabling access to news sources and semantic news search capabilities.

Setup Requirements

  • ⚠️Requires Python 3.7+ for the Python package.
  • ⚠️Requires Qt and CMake 3.2+ for the C++ client.
  • ⚠️Access to the World News API requires an API key, which might entail registration or a paid plan.
  • ⚠️Requires internet access to connect to the external World News API.
Verified SafeView Analysis
The project is an API client library, shifting primary security responsibilities (API key storage, input sanitization) to the consuming application. API keys are passed in query parameters or headers, which is standard but requires HTTPS for confidentiality. The C++ client includes a local TCP ReplyServer for OAuth redirect handling, which, while standard, needs careful implementation to prevent local network vulnerabilities. URL parameter substitution uses regular expressions, which could be a ReDoS vector if server variable values were untrusted, but they are typically fixed configuration. There are no obvious hardcoded secrets or malicious patterns in the provided truncated code.
Updated: 2025-12-13GitHub→
0
0
Medium Cost
AbdelilahOu icon

DBMcp

by AbdelilahOu

Sec3

This MCP server provides AI assistants with a standardized interface to connect, query, and manage various SQL databases, enabling data interaction and schema exploration.

Setup Requirements

  • ⚠️Requires direct network access to the configured PostgreSQL and MySQL databases from the server's host environment.
  • ⚠️A 'connections.json' configuration file, containing database connection details and credentials, must be provided to the server.
  • ⚠️To build the server from source, a Go development environment (Go 1.18+ recommended) is required.
Review RequiredView Analysis
CRITICAL: Database credentials (usernames and passwords) are hardcoded directly into 'connections.json' as plain text. This file is likely committed to the repository, representing a severe exposure of sensitive information. The 'execute_query' tool allows direct execution of arbitrary SQL (INSERT, UPDATE, DELETE, CREATE, ALTER, DROP) provided by an AI, with only a weak blacklist for specific dangerous operations (like 'DROP DATABASE' or 'TRUNCATE'). This makes the server highly vulnerable to SQL injection, unauthorized data manipulation, and schema destruction if the AI's input is compromised or crafted maliciously. There is no explicit authentication or authorization layer implemented for AI clients interacting with the server's tools, relying solely on the underlying MCP transport's security (which is 'stdio' for local clients in the provided configuration, but can change).
Updated: 2025-12-01GitHub→
0
0
Low Cost
anandramkumar2020 icon

mcp-101

by anandramkumar2020

Sec9

An educational MCP (Model Context Protocol) server demonstrating fundamental concepts and providing simple calculator and API bridge tools for AI models.

Setup Requirements

  • ⚠️Requires Python 3.10 or higher
  • ⚠️Requires `uv` package manager for dependency management
Verified SafeView Analysis
No 'eval' or similar dynamic code execution found in the server or tool implementation files. Tool calls are explicitly routed based on string matching (`if name == "x_add"`) in `server.py`, preventing arbitrary function calls. The `x_api_demo` function explicitly uses a simulated response, avoiding direct external network calls within this demo implementation. Input parameters for calculator functions expect integers, and division by zero is handled gracefully.
Updated: 2025-11-23GitHub→
0
0
Low Cost
Ryno-Crypto-Mining-Services icon

braiins-os-mcp-server

by Ryno-Crypto-Mining-Services

Sec7

AI-assisted management and monitoring of Braiins OS ASIC mining operations via Model Context Protocol (MCP), enabling AI agents like Claude to interact with miner firmware.

Setup Requirements

  • ⚠️Requires Node.js 20.x LTS.
  • ⚠️Docker and Docker Compose are listed as prerequisites, but recent internal reviews note deployment infrastructure (Dockerfile, docker-compose.yml, K8s configs) is incomplete.
  • ⚠️Relies on a non-native cron scheduling workaround, as the Braiins OS API lacks native cron support for power schedules. This requires an application-level scheduler or external system cron.
  • ⚠️A recent internal review (2026-01-13) notes 9 failing tests and broken test coverage reporting, which require investigation for full stability.
Verified SafeView Analysis
The project demonstrates strong intent and implementation of security best practices, including TLS 1.3 for transport, JWT for authentication, role-based access control (RBAC), AES-256 encryption for data at rest, comprehensive audit logging, and masking of sensitive data in logs. It also uses `helmet` for HTTP security. However, internal documentation (PRODUCTION_READINESS_CHECKLIST.md) indicates the OWASP Top 10 security audit is 0% executed, suggesting formal verification and hardening is still in progress. No critical immediate vulnerabilities (e.g., 'eval', hardcoded critical secrets) were found.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
bassa01 icon
Sec8

Provides a Model Context Protocol (MCP) server that connects to Google Cloud services to offer context and tools for interacting with various Google Cloud resources.

Setup Requirements

  • ⚠️Requires Node.js 24.11.0 or newer.
  • ⚠️Requires pnpm 10.21.0+ via Corepack.
  • ⚠️Requires Google Cloud authentication (e.g., `GOOGLE_APPLICATION_CREDENTIALS` or `gcloud` ADC) configured to access GCP services.
  • ⚠️The `gcloud-run-read-command` tool requires the Google Cloud CLI (`gcloud`) to be installed and available on the system's PATH.
Verified SafeView Analysis
The server integrates with the Google Cloud CLI (`gcloud`), which is inherently a high-risk operation. However, the codebase implements robust guardrails to mitigate this risk, including: 1. A strict read-only policy for `gcloud` commands, blocking mutable verbs (e.g., `create`, `delete`, `update`) and sensitive prefixes (e.g., `iam`, `secret-manager`, `kms`). 2. Enforcement of service account identity for `gcloud` commands, preventing execution with personal credentials. 3. Comprehensive read-only assertion for SQL queries in BigQuery and Spanner, explicitly denying destructive statements (`INSERT`, `UPDATE`, `DELETE`, `DROP`, `ALTER`, `CREATE`). 4. Log redaction policies to scrub sensitive data (IPs, user identifiers, request bodies) from Logging outputs unless explicitly authorized by roles. 5. General input sanitization and rate-limiting via an `McpSecurityValidator`. While the `gcloud` CLI integration remains a potential vector, the implemented policies significantly reduce its exploitability. No obvious hardcoded secrets or `eval` misuse were found.
Updated: 2025-11-19GitHub→
PreviousPage 480 of 760Next