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
Medium Cost
thsmale icon
Sec8

An LLM agent server designed to provide access to the BibleGet I/O API for quoting Sacred Scriptures via the Model Context Protocol (MCP).

Setup Requirements

  • ⚠️Requires 'uv' (ultrafast Python package installer and runner) to be installed.
  • ⚠️Requires 'OPENAI_API_KEY' environment variable to be set for the client (which interacts with this server) to function.
Verified SafeView Analysis
The server fetches its OpenAPI specification from an external GitHub URL (raw.githubusercontent.com) at startup. A compromise of this external source could lead to changes in the tools presented to the LLM. However, no 'eval' or direct obfuscation is found, and sensitive configurations are handled via environment variables.
Updated: 2025-11-19GitHub→
0
0
Low Cost
junimnjw icon

mcp-server-test

by junimnjw

Sec1

This project aims to simulate or test the Minecraft Protocol (MCP) for development and compatibility verification.

Review RequiredView Analysis
Source code was not provided for analysis. Therefore, a security audit could not be performed to check for 'eval', obfuscation, network risks, hardcoded secrets, or malicious patterns. Running unverified code carries significant risks, including potential for malicious activity, data breaches, or system compromise. It is considered unsafe without a full code review.
Updated: 2025-11-22GitHub→
0
0
Medium Cost
inmydata icon

AgentMCP

by inmydata

Sec5

Provides AI agents with structured and conversational data querying, plus calendar functionalities, via a Model Context Protocol (MCP) server.

Setup Requirements

  • ⚠️Requires `inmydata` API Key, Tenant, and Calendar Name.
  • ⚠️Python 3.11 is the minimum runtime.
  • ⚠️`MCP_DUCKDB_LOCATION` environment variable must be configured for DuckDB storage.
  • ⚠️If using OAuth with Personal Access Tokens (PATs), `INMYDATA_INTROSPECTION_CLIENT_ID` and `INMYDATA_INTROSPECTION_CLIENT_SECRET` are required.
Review RequiredView Analysis
The `query_results_fast` tool allows direct execution of user-provided SQL against a local DuckDB database. While confined to a local file, this presents a significant SQL injection risk if the AI agent or end-user provides malicious SQL. The server supports both OAuth (with token introspection for PATs) and API key authentication, handling credentials via environment variables or HTTP headers. It requires client ID/secret for OAuth introspection, which are read from environment variables. Proper HTTPS is recommended for remote deployments.
Updated: 2025-12-08GitHub→
0
0
Medium Cost
civicteam icon

bill-mcp-server

by civicteam

Sec8

An MCP server for Bill.com Spend & Expense API integration, enabling AI assistants to manage financial operations related to company spending, budgets, reimbursements, transactions, cards, and users.

Setup Requirements

  • ⚠️Requires a Bill.com Spend & Expense API token with appropriate permissions, which must be obtained from a Bill.com developer account.
  • ⚠️Requires Node.js 18.0.0 or higher runtime environment.
  • ⚠️The environment variables `BILL_API_TOKEN` (required) and `BILL_ENVIRONMENT` (optional, defaults to `production`) must be set for the server to function.
Verified SafeView Analysis
Authentication relies on a `BILL_API_TOKEN` environment variable, which is a secure practice to avoid hardcoding credentials. The server uses the native `fetch` API for making requests, passing the API token directly in the 'apiToken' header. IDs are directly interpolated into URL paths for `GET` requests, which is generally safe when the IDs are expected to be simple strings. Error handling catches general exceptions and returns a generic 'Unknown error occurred' message, preventing the exposure of detailed internal server errors or stack traces to external callers. No 'eval' or explicit obfuscation techniques were detected in the provided source code. It relies on the security of the underlying Bill.com API for input validation and access control beyond what this server explicitly implements.
Updated: 2026-01-13GitHub→
0
0
Low Cost
gtrevize icon

mcp-network

by gtrevize

Sec9

A comprehensive, secure MCP (Model Context Protocol) server for AI agents to perform remote network testing, validation, and diagnostics.

Setup Requirements

  • ⚠️Requires Node.js >= 18.0.0.
  • ⚠️Critical environment variables (JWT_SECRET, AUTH_TOKEN) must be set and not contain 'CHANGEME' default values for server startup or full tool functionality.
  • ⚠️Prerequisite system tools (e.g., ping, traceroute, dig, whois, nmap, tcpdump, iperf3) must be installed on the host, and some (nmap, tcpdump) may require elevated privileges (sudo or capabilities).
Verified SafeView Analysis
The server implements robust security measures including JWT authentication with RBAC, comprehensive input validation, and explicit anti-jailbreaking guardrails. It actively checks for malicious patterns like shell injection, path traversal, and dangerous commands. `child_process.exec` and `spawn` are used for tools but with validated inputs and timeouts. The `tls-test` tool intentionally uses `rejectUnauthorized: false` for diagnostic purposes, which is documented and justified. A critical initial setup step requires changing the `JWT_SECRET` from a default `CHANGEME` value, which is actively validated at startup.
Updated: 2025-11-18GitHub→
0
0
Medium Cost
undiabler icon

polymarket-mcp

by undiabler

Sec9

Provides an in-memory cache and tools for LLM agents to explore, search, and filter Polymarket event data for agentic integrations.

Setup Requirements

  • ⚠️Requires Docker for deployment
  • ⚠️Requires `MCP_BEARER_TOKEN` environment variable for server security
  • ⚠️Recommended: Docker volume for persistent cache (`--volume polycache:/app/data`) to avoid slow startup on restarts
Verified SafeView Analysis
The server uses a required `MCP_BEARER_TOKEN` environment variable for authentication, preventing hardcoded secrets. It employs an `ArgumentSanitizerMiddleware` to strip unknown fields from tool arguments, mitigating potential injection risks from MCP clients. External API calls use `httpx` with timeouts and rate limiting, reducing network attack surface. No `eval` or other dangerous dynamic code execution patterns were found. The primary risk is the user choosing a weak `MCP_BEARER_TOKEN`.
Updated: 2025-12-19GitHub→
0
0
Low Cost

egw_writings_mcp_server

by pythondev-pro

Sec8

Provides access to Ellen G. White writings data via MCP and HTTP APIs, including search, content retrieval, and PDF generation.

Setup Requirements

  • ⚠️Requires local download of EGW database (large file size)
  • ⚠️Multi-tenant server (`pyforge-multi-tenant-mcp-server`) requires careful management of session lifecycles and potential file system access if not configured securely.
  • ⚠️The `admin_local_server` tool uses a default admin password ('admin18401844') which should ideally be changed or managed via environment variables for production.
Verified SafeView Analysis
The system uses local processing and does not inherently require cloud services after initial database download. However, the pyforge-multi-tenant-mcp-server component relies on `GITHUB_PAT` for Git operations, which, if exposed, could be a risk. Other components seem sandboxed or operate on local data. The 'admin_local_server' tool requires a password, which is good, but its default password could be stronger.
Updated: 2025-11-21GitHub→
0
0
High Cost
hrayleung icon

Cocode

by hrayleung

Sec9

Provides a high-performance MCP server for semantic codebase search, featuring real-time incremental indexing and intelligent ranking.

Setup Requirements

  • ⚠️Requires PostgreSQL with the `pgvector` extension installed and enabled (the server will attempt to enable `pgcrypto` too).
  • ⚠️Requires an API key for at least one embedding provider (OpenAI, Jina, or Mistral), which are typically paid services.
  • ⚠️Requires Python 3.10+.
Verified SafeView Analysis
The codebase demonstrates robust security practices. It explicitly uses environment variables for API keys and database credentials, preventing hardcoded secrets. Database interactions use parameterized queries with `psycopg.sql.SQL` to mitigate SQL injection risks. Path validation is implemented to prevent directory traversal attacks and ensure proper file access. External API calls (Jina, Mistral, OpenAI, Cohere) are rate-limited and have timeouts configured to prevent resource exhaustion. There are no obvious `eval` or `exec` calls, and it uses well-known, peer-reviewed libraries. Overall, the risk of common vulnerabilities appears low for a local MCP server.
Updated: 2026-01-19GitHub→
0
0
Low Cost
isakskogstad icon

personupplysning-mcp

by isakskogstad

Sec9

Provides a Model Context Protocol (MCP) server for accessing Swedish company data and financial reports from Bolagsverket API, with local caching in Supabase.

Setup Requirements

  • ⚠️Requires a Supabase project (Pro plan recommended for 3.1GB dataset, $25/month). Free tier pauses and is too small for the dataset.
  • ⚠️Requires Bolagsverket API credentials (Client ID and Client Secret). The free tier is limited to 1000 API calls/month.
  • ⚠️Requires initial import of 1.85M Swedish company records from Parquet files into Supabase, which takes approximately 46 minutes (`npm run db:import`).
Verified SafeView Analysis
The server employs comprehensive input validation using Zod schemas with regex patterns to prevent XSS and SQL injection. Error responses are structured, and stack traces are suppressed in production to prevent information leakage. Environment variables are validated on startup to ensure critical configurations are present. OAuth2 token management includes caching and refresh mechanisms. CORS is configured with `origin: '*'` which is typical for public APIs but could be narrowed if clients are known. An architectural inconsistency exists regarding the primary MCP transport (HTTP JSON-RPC vs. SSE) and endpoints (`/mcp` vs `/sse` and `/messages`), which might lead to confusion but doesn't inherently pose a security vulnerability.
Updated: 2025-12-01GitHub→
0
0
Medium Cost
alestmej icon

mcp-python

by alestmej

Sec9

An AI agent leveraging LangChain to interact with a Minecraft Protocol (MCP) environment.

Setup Requirements

  • ⚠️Requires Python 3.14 or newer (as per `pyproject.toml`, noting Python 3.14 is not yet released).
  • ⚠️Requires `TOKEN` environment variable to be set.
Verified SafeView Analysis
The provided `main.py` is simple and secure, using environment variables for secrets (`TOKEN`) and avoiding dangerous functions like `eval`. Dependencies (`langchain-openai`, `langchain-mcp-adapters`) will involve external API calls, which introduce typical API key security and prompt injection considerations for AI applications, but these are not directly present in the audited code snippet itself.
Updated: 2025-12-06GitHub→
0
0
Medium Cost
george7979 icon

gpt-mcp-server

by george7979

Sec9

An MCP server that provides OpenAI GPT capabilities to Claude Code and other MCP clients, utilizing the OpenAI Responses API for advanced models and features.

Setup Requirements

  • ⚠️Requires an OpenAI API Key (Paid) configured as an environment variable (OPENAI_API_KEY).
  • ⚠️Requires Node.js version 18 or higher.
  • ⚠️The configured GPT model (via GPT_MODEL env var) is validated at startup; if invalid, it falls back to 'gpt-5.1-codex' with a warning.
Verified SafeView Analysis
The server follows good security practices by requiring the OpenAI API key via an environment variable (OPENAI_API_KEY) and performing input validation using Zod's `.strict()` schemas. It communicates via standard I/O (stdio), limiting direct network exposure. Error handling includes actionable messages without exposing sensitive internal details. No 'eval' or obfuscation is present. The primary security consideration is the reliance on a third-party OpenAI API.
Updated: 2025-11-25GitHub→
0
0
Medium Cost
MyMindVentures icon

MMC_MCP_Server_Bridge

by MyMindVentures

Sec1

Acts as a unified HTTP/GraphQL gateway for a multitude of Model Context Protocol (MCP) servers, enabling agentic workflow orchestration and cross-service communication.

Setup Requirements

  • ⚠️Hardcoded API keys and connection strings in 'server.js' must be removed and securely managed via environment variables before deployment.
  • ⚠️Extensive environment variable configuration is required for all 25+ integrated MCP servers to function correctly.
  • ⚠️The Railway CLI and an authenticated session are prerequisites for the 'railway-mcp' server, potentially requiring manual setup outside the Docker container.
Review RequiredView Analysis
CRITICAL: The system contains multiple instances of hardcoded API keys and connection strings (e.g., N8N_API_KEY, Railway API Key, MongoDB Connection String, Raindrop credentials) directly within 'server.js'. Although environment variables are used as a fallback for some, the MongoDB connection string is explicitly passed as a direct argument to a child process, making it unconditionally exposed in the source code. This poses a critical security vulnerability, as these credentials would be publicly visible to anyone with access to the codebase.
Updated: 2025-12-03GitHub→
PreviousPage 636 of 713Next