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
krisrowe icon

monarch-access

by krisrowe

Sec9

Provides AI assistants and agentic tools with secure access to Monarch Money financial data for querying and taking actions.

Setup Requirements

  • ⚠️Requires Python 3.10+.
  • ⚠️Requires a Monarch Money account and manual extraction of a browser session token via DevTools.
  • ⚠️Requires Docker to run the server in a containerized environment.
  • ⚠️Requires an MCP-compatible client (e.g., Claude Desktop, Gemini CLI) to utilize the server.
Verified SafeView Analysis
The server uses standard `aiohttp` for GraphQL API calls and `json` for data handling, with no use of `eval` or similar dangerous functions. Authentication relies on a Monarch Money browser session token, which users must manually extract; the server stores this token in `~/.config/monarch/token` by default or can receive it via an environment variable. Explicit warnings are provided to not commit tokens to version control. The Docker setup for HTTP transport maps port 8000, and the `uvicorn` server inside typically binds to `0.0.0.0`, making it accessible on the host machine. While the `MCP-SERVER.md` mentions 'Local only' by default, binding to `0.0.0.0` in Docker could technically allow external access if not properly firewalled, though its primary use is local AI agent integration.
Updated: 2025-12-22GitHub→
0
0
Medium Cost
Dhana009 icon

graphiti

by Dhana009

Sec8

A Micro-Agent Coordination Protocol (MCP) server for managing and querying a Neo4j-based knowledge graph, featuring LLM-powered entity/relationship extraction and semantic search.

Setup Requirements

  • ⚠️Requires OpenAI API Key (Paid Service)
  • ⚠️Requires a running Neo4j database (Docker recommended for local setup)
  • ⚠️Requires Python 3.10 or higher
  • ⚠️Uses 'uv' package manager in example run command (can be replaced with 'python')
Verified SafeView Analysis
The project demonstrates strong input validation via `src/validation.py` for all entity and relationship fields, significantly mitigating common injection risks. Neo4j queries use parameterized statements (`$param`), preventing Cypher injection. LLM output (JSON) is parsed and its structure is validated before further processing, avoiding direct code execution. The `OPENAI_API_KEY` is correctly handled as an environment variable, not hardcoded. Default Neo4j credentials are for a local development setup, not production. Potential risks largely stem from LLM vulnerabilities (e.g., advanced prompt injection) or misconfiguration in a production environment, rather than direct code-level flaws.
Updated: 2025-11-24GitHub→
0
0
Medium Cost
RetroFury07 icon

My-Digital-Twin

by RetroFury07

Sec7

A comprehensive digital twin platform leveraging Retrieval-Augmented Generation (RAG) with LLM-enhanced queries and responses for professional profile querying and interview preparation, featuring a web-based chat interface and monitoring.

Setup Requirements

  • ⚠️Requires Groq API Key, which has a generous free tier but can incur costs if usage exceeds limits.
  • ⚠️Requires an Upstash Vector index with 1536 dimensions, and embeddings must be generated (via `embed_digitaltwin.py`) using an OpenAI API key or a local `sentence-transformers` model, which is then padded to match the 1536 dimension.
  • ⚠️The TypeScript `lib/vector-db.ts` implementations for Upstash and Pinecone search are commented out and currently fall back to mock data, requiring further implementation for live vector database integration for search queries.
Verified SafeView Analysis
The Python FastAPI servers and the raw Python BaseHTTPRequestHandler serverless function (`api/rag.py` and `mydigitaltwin/api/rag.py`) explicitly set `Access-Control-Allow-Origin: *`. While suitable for development, this broad CORS policy could be a security risk in a production environment if the API were to handle or expose more sensitive user data beyond the public-facing profile or be integrated into a larger system without proper origin restrictions. API keys are loaded from environment variables, which is good practice. No 'eval' or obvious malicious patterns were found in the provided code snippets.
Updated: 2025-11-29GitHub→
0
0
Low Cost
toshio999 icon
Sec3

Enables local LLM clients (like Claude Desktop) to query on-premise Symfoware databases using natural language.

Setup Requirements

  • ⚠️Requires pre-installed Fujitsu Symfoware Client with .NET Data Provider and its assemblies registered in the GAC.
  • ⚠️Requires Windows OS and .NET Framework 4.8 runtime environment.
  • ⚠️Requires manual configuration of Claude Desktop's `claude_desktop_config.json` to specify the server executable and connection details.
Review RequiredView Analysis
The primary security concern is the `query_database` function, which allows LLM-generated SQL queries to be executed directly against the Symfoware database. While a `SYMFOWARE_SELECT_ONLY` flag exists to restrict operations to `SELECT` statements, it does not prevent SQL injection attacks that could lead to data exfiltration or denial of service through complex queries. The provided documentation does not detail any robust SQL sanitization, parameterization, or allowlisting mechanisms to mitigate these risks. Without the actual C# source code for query execution, a significant risk of SQL injection must be assumed. No 'eval' or obfuscation is apparent, and hardcoded secrets are not visible as the connection string is expected to be an environment variable.
Updated: 2026-01-17GitHub→
0
0
Low Cost
Traia-IO icon
Sec8

Interact with CoinmarketCap v106 API via an AI-friendly Model Context Protocol (MCP) server, supporting both authenticated access with client API keys and pay-per-use via a blockchain-based HTTP 402 payment protocol.

Setup Requirements

  • ⚠️Requires a CoinmarketCap API Key (`COINMARKET_V106_API_KEY`), which can be client-provided (free mode) or server-provided (paid mode).
  • ⚠️Docker (or Python 3.12+ with `uv` for manual installation) is required.
  • ⚠️Requires Ethereum wallet details (`SERVER_ADDRESS`, `MCP_OPERATOR_PRIVATE_KEY`, `MCP_OPERATOR_ADDRESS`) for the D402 payment protocol; a local script can auto-generate these for development but relies on the `web3` Python library.
Verified SafeView Analysis
The server uses `allow_origins=["*"]` with `allow_credentials=True` in its CORS configuration, which is broad but common for public APIs intended for various client integrations like LLMs. It heavily relies on the `traia_iatp.d402` library for HTTP 402 payment protocol and blockchain verification; the security posture is significantly dependent on the robustness and audit status of this external library. The `run_local_docker.sh` script dynamically generates Ethereum keys (`SERVER_ADDRESS`, `MCP_OPERATOR_PRIVATE_KEY`, `MCP_OPERATOR_ADDRESS`) for local development if not present, which is a convenient workaround but adds complexity to the local setup process. API calls to CoinmarketCap use hardcoded URLs and include timeouts, reducing risk from arbitrary external calls. No direct `eval`, code obfuscation, or overtly malicious patterns were identified in the provided source.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
emrgcl icon
Sec9

An AI-powered toolkit for making informed Azure region selection decisions by combining network latency testing and pricing analysis.

Setup Requirements

  • ⚠️Requires Python 3.10+.
  • ⚠️Azure CLI must be installed and authenticated (`az login`).
  • ⚠️Your Azure account needs permissions to create and delete Azure Resource Groups and Storage Accounts.
Verified SafeView Analysis
The server uses DefaultAzureCredential for secure Azure authentication, avoiding hardcoded secrets. It employs Pydantic for input validation, minimizing injection risks. Temporary Azure resources (resource groups and storage accounts) are created for latency testing and include automatic, robust cleanup mechanisms, even on cancellation. Requires Azure permissions to create/delete these resources.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
ashutoshkj icon

my-mcp-server

by ashutoshkj

Sec1

A server implementing the Minecraft Protocol, allowing clients to connect and interact within a game world.

Review RequiredView Analysis
CRITICAL: A comprehensive security audit cannot be performed as no source code was provided. The server's nature suggests it handles network connections and potentially game logic, making security a high priority. Without code review, it's impossible to check for vulnerabilities like 'eval' usage, hardcoded secrets, network attack vectors, or other malicious patterns. Running this server without reviewing its source code is highly risky.
Updated: 2025-11-30GitHub→
0
0
Low Cost
dylanmarriner icon

KAIZA-MCP-server

by dylanmarriner

Sec9

Enterprise governance gateway for AI-driven development, transforming unconstrained AI agents into governed execution authorities.

Setup Requirements

  • ⚠️Requires Node.js 18+ and npm 9+.
  • ⚠️Mandatory environment variable `KAIZA_BOOTSTRAP_SECRET` must be set for initial plan creation.
  • ⚠️Requires an MCP-compatible client (e.g., Windsurf, Claude in Cline) for interaction, as it communicates over standard I/O.
  • ⚠️The `read_prompt` tool must be called first in any session before `write_file` operations are permitted.
Verified SafeView Analysis
The KAIZA MCP server is designed with a strong 'security-first' and 'fail-closed' philosophy. It implements a zero-trust execution model, plan-based authorization, cryptographic audit trails, and extensive write-time policy enforcement (stub detection, language-specific rules, intent co-requirement). Path traversal is strictly prevented, and all write operations are subject to preflight checks and atomic reverts upon failure. All secrets are expected via environment variables. Communication is primarily over stdio, reducing network attack surface. Comprehensive adversarial testing and remediations are documented.
Updated: 2026-01-19GitHub→
0
0
Low Cost
nidajumail icon

mcp_server

by nidajumail

Sec1

Unable to determine the specific use case as no source code was provided for analysis.

Review RequiredView Analysis
No source code was provided for analysis. Therefore, a comprehensive security audit cannot be performed. It is impossible to identify 'eval' usage, obfuscation, network risks, hardcoded secrets, or any malicious patterns without access to the code. Running software without code review is inherently risky.
Updated: 2025-11-25GitHub→
0
0
Medium Cost
kalivemularajendra icon

BigQuery-Assistant

by kalivemularajendra

Sec9

A FastMCP-powered BigQuery server that provides intelligent data discovery and analytics capabilities through specialized AI agents.

Setup Requirements

  • ⚠️Requires a Google Cloud BigQuery project ID.
  • ⚠️Requires Google Cloud authentication (service account, ADC, or gcloud login).
  • ⚠️Requires installation of Python dependencies: fastmcp, google-cloud-bigquery, python-dotenv, google-adk.
Verified SafeView Analysis
The server uses environment variables for sensitive BigQuery project ID, location, and service account key file, which is good practice. It leverages Google Cloud's official `google-cloud-bigquery` and `google.auth` libraries for authentication, supporting service account key files and Application Default Credentials. There are no obvious hardcoded secrets, `eval` calls, or obfuscation. Logging of the key file path is intentionally avoided. Network communication uses standard HTTP/SSE.
Updated: 2025-12-01GitHub→
0
0
Low Cost
ViperJuice icon

pmcp

by ViperJuice

Sec8

Acts as a single MCP server for Claude Code to reduce context bloat by providing on-demand tool discovery and dynamic server provisioning.

Setup Requirements

  • ⚠️Requires Python 3.10+.
  • ⚠️Node.js (with npm/npx) and/or a Python package manager (like `uv` or `pip`) must be available on the system to install and run many downstream MCP servers.
  • ⚠️For LLM-enhanced capability matching and summarization (optional `pmcp[llm]` feature), a `GROQ_API_KEY` is required.
  • ⚠️Many downstream MCP servers (e.g., GitHub, Brave Search, Notion, Postgres) require their own specific API keys, which must be set as environment variables.
Verified SafeView Analysis
The server's core function involves spawning and managing external processes (other MCP servers or CLIs) via `asyncio.create_subprocess_exec`. While this is necessary for its design, it inherently relies on the trustworthiness of the manifest (`manifest.yaml`) and the integrity of installed downstream packages (npm/PyPI). The `PolicyManager` is implemented to mitigate risks through allow/deny lists for servers, tools, resources, and prompts, as well as output size caps and secret redaction. Crucially, a singleton lock and self-reference filtering are in place to prevent recursive spawning. No obvious `eval` or obfuscation was found, and hardcoded secrets are avoided by relying on environment variables. The primary security consideration is the vigilance required in managing and auditing the list of allowed external tools.
Updated: 2026-01-17GitHub→
0
0
Low Cost
carlosGalisteo icon

mcp_server_example

by carlosGalisteo

Sec9

This server exposes simple Python functions as API endpoints using the FastMCP framework, providing a demonstration of a microservice.

Setup Requirements

  • ⚠️Requires Python version 3.12 or higher.
  • ⚠️Relies on the 'mcp[cli]' library, which needs to be installed.
Verified SafeView Analysis
The provided source code is minimal and implements a simple addition function, which presents no inherent security risks like 'eval', obfuscation, hardcoded secrets, or malicious patterns. The overall security profile would heavily depend on the underlying 'mcp' framework and its implementation, which is not fully detailed here.
Updated: 2025-12-03GitHub→
PreviousPage 331 of 760Next