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
Low Cost
AutoByteus icon

autobyteus_mcps

by AutoByteus

Sec7

The server automates the creation and modification of PowerPoint presentations and combined images using only image inputs, scaling them to fit slides or composite images without cropping.

Setup Requirements

  • ⚠️Python 3.10+ required.
  • ⚠️Requires `python-pptx` and `Pillow` libraries (installed via `uv` or `pip`).
  • ⚠️Path handling relies on `AUTOBYTEUS_AGENT_WORKSPACE` environment variable for relative paths, otherwise defaults to server CWD. Absolute paths are recommended for predictability.
Verified SafeView Analysis
Direct use of user-provided output paths in file save operations (e.g., `prs.save`, `new_im.save`) could allow writing to arbitrary file system locations if the output path is not properly sanitized or constrained by the calling agent. Relying on `AUTOBYTEUS_AGENT_WORKSPACE` for CWD changes means the server's working directory can be influenced by the environment, which should be set securely. Image processing with `Pillow` can be vulnerable to resource exhaustion from malformed image inputs.
Updated: 2025-12-24GitHub→
0
0
Medium Cost
xenoxavier icon

Indicators

by xenoxavier

Sec4

This server provides programmatic access to analyze TradingView Pine Script indicators via the Model Context Protocol (MCP).

Setup Requirements

  • ⚠️Requires Node.js >=18 to run.
  • ⚠️Indicator files must be located in the server's working directory or a specifically mounted/configured path.
  • ⚠️Designed to work with a Model Context Protocol (MCP) client, typically Claude Desktop.
Review RequiredView Analysis
The server uses `fs.readdirSync(".")` and `fs.readFileSync(filename, 'utf-8')` to access indicator files. The `analyze_indicator`, `search_indicators`, and `extract_functions` tools take `indicatorName` or `searchTerm` as direct string inputs. While `getIndicatorFiles` attempts to filter out non-indicator files based on common patterns (e.g., `file.startsWith('.')`, `file.includes('package')`), this filtering is regex-based and applied to filenames in the current directory. This approach is vulnerable to path traversal (e.g., `../../sensitive_file.txt`) or cleverly named files that could bypass the filters, potentially allowing an attacker to read arbitrary files within the server's working directory or mounted volumes. The Docker configuration mitigates write risks by mounting the volume as read-only, but reading sensitive files (e.g., configuration, logs) within the mounted context remains a risk if the filters are imperfect or bypassed. There are no obvious signs of 'eval' or hardcoded secrets.
Updated: 2025-11-25GitHub→
0
0
Low Cost
jinyiabc icon

ibkr-mcp

by jinyiabc

Sec9

An MCP server that enables Claude AI to interact with Interactive Brokers trading accounts for market data, account management, and trade execution.

Setup Requirements

  • ⚠️Requires an active Interactive Brokers account (paper or live).
  • ⚠️Requires Interactive Brokers Trader Workstation (TWS) or IB Gateway to be installed, running, and manually configured for API access.
  • ⚠️Requires Python 3.8+ to be installed and added to the system PATH.
  • ⚠️Manual configuration of Claude Desktop's `claude_desktop_config.json` is necessary for integration.
Verified SafeView Analysis
The server uses a local-only Model Context Protocol (MCP) server, communicating via standard I/O (stdio), which limits network exposure. The connection to Interactive Brokers TWS/Gateway is also configured for localhost (127.0.0.1) by default, further enhancing local security. No hardcoded credentials or API keys were found. The project strongly emphasizes starting with paper trading, understanding the risks of live trading, and securely configuring TWS/Gateway API access (e.g., unchecking 'Read-Only API' for order placement). The primary security risk lies in the user's operational choices, particularly enabling live trading for automated order placement without sufficient caution, which is clearly warned against in the documentation.
Updated: 2026-01-19GitHub→
0
0
Low Cost
bioanywhere icon

manana2

by bioanywhere

Sec8

Provides client libraries to access and consume news data from the World News API for various applications.

Setup Requirements

  • ⚠️Requires an API key from World News API (paid service likely).
  • ⚠️Requires Python 3.7+ for Python client library.
  • ⚠️Requires network access to `https://api.worldnewsapi.com`.
Verified SafeView Analysis
The repository contains client libraries for various programming languages. These libraries are generally safe as they are designed to consume an external API. No explicit malicious patterns or obfuscation were found. API keys are expected to be provided by the user and are not hardcoded within the client source. A minor potential concern in the C++ client's OAuth `ReplyServer` is that it listens on `QHostAddress::Any` (all network interfaces) for callbacks on port 9999, which ideally should be restricted to `QHostAddress::LocalHost` in typical client-side OAuth flows to prevent external access, although this is usually a transient local service.
Updated: 2025-12-13GitHub→
0
0
Low Cost
HuyDoan-2112 icon

setup-mcpserver-basic

by HuyDoan-2112

Sec3

Provides a foundational server implementation for the Minecraft Protocol, suitable for custom game logic, mini-games, or server administration.

Setup Requirements

  • ⚠️Requires a specific runtime environment (e.g., Java, Node.js, Python), which is unknown without the source code.
  • ⚠️Manual configuration of network ports may be required for external access.
  • ⚠️No setup instructions are available due to the missing README and inaccessible source code.
Review RequiredView Analysis
A detailed security audit is impossible as the full source code was not provided for analysis. Without access to the code, common server vulnerabilities such as unvalidated input, insecure network handling, or potential hardcoded credentials cannot be assessed. It is unsafe to run without a thorough code review by a trusted party.
Updated: 2025-12-02GitHub→
0
0
Low Cost
rykhalskyi icon

ragatouille

by rykhalskyi

Sec8

A local RAG and MCP server solution for building personalized knowledge bases from diverse data sources.

Setup Requirements

  • ⚠️Docker or Podman required for easy deployment.
  • ⚠️Project is in 'pre-production' status, implying potential instability.
  • ⚠️Local disk space will be consumed by ChromaDB for vector data.
Verified SafeView Analysis
The server exposes ports 4301 (FastAPI), 4302 (FastMCP), and 4300 (frontend) which bind to 0.0.0.0 in the Docker setup. While standard for local external access, this means they are accessible from any network interface, which could be a risk if deployed on a public server without proper firewall rules. No 'eval' or other obvious malicious patterns found. No hardcoded sensitive API keys or credentials. CORS is configured for localhost by default, which is safe. The project is noted as 'pre-production' which suggests potential for instability or undiscovered issues.
Updated: 2026-01-18GitHub→
0
0
Low Cost
theREDspace icon

mcp-server-example

by theREDspace

Sec4

Provides a basic demonstration of server functionality, likely for educational purposes or as a starting point for network application development.

Setup Requirements

  • ⚠️Requires Node.js runtime environment to be installed.
  • ⚠️Requires a free network port (e.g., 3000, 8080) for the server to listen on.
  • ⚠️Project dependencies must be installed (e.g., `npm install`).
Review RequiredView Analysis
Source code for the server was not provided, preventing a comprehensive security audit. As an 'example' server, it is likely not hardened for production environments and may contain default insecure configurations or unaddressed vulnerabilities. Running unknown server code carries inherent risks. Specific checks for 'eval', obfuscation, hardcoded secrets, and network risks could not be performed.
Updated: 2026-01-18GitHub→
0
0
High Cost
hpatel-solusoft icon

mcp-server-poc

by hpatel-solusoft

Sec8

An AI-driven system for autonomous processing of insurance claims (motor and health) by extracting document information, orchestrating workflows in an external Case360 system, and storing records locally.

Setup Requirements

  • ⚠️Requires an OpenAI API Key (Paid API) for the Python agent.
  • ⚠️Requires access to a running instance of 'Case360' or a compatible SOAP web service, including its URL, username, and password for the Java server.
  • ⚠️The Streamlit dashboard (dashboard.py) hardcodes the SQLite database file path ('D:\Solusoft\AI\MCP\my_work\claims-mcp-server\src\data\claims.db') which must be updated for other environments.
Verified SafeView Analysis
The Java server uses `@Value` for configuration (Case360 credentials) which is good practice. SQL queries are parameterized, preventing injection. The Python client uses `load_dotenv()` for API keys, which is standard. No direct `eval` or obvious obfuscation. The system integrates with an external Case360 system via SOAP with basic authentication, which needs to be properly secured at the endpoint. Base64 decoding for document upload is present; extreme untrusted input could pose a DoS risk, but it's typically an internal tool.
Updated: 2026-01-16GitHub→
0
0
High Cost
jakvbs icon

claude-mcp-rs

by jakvbs

Sec9

A high-performance Rust-based MCP server that integrates with the Claude CLI for AI-assisted coding tasks, enabling multi-turn conversations and configurable execution.

Setup Requirements

  • ⚠️Requires Claude CLI installed and configured locally on the system
  • ⚠️Requires an MCP client (e.g., Claude Code) for practical interaction
Verified SafeView Analysis
The server uses Rust's `tokio::process::Command` which handles arguments safely, preventing shell injection. Inputs are validated (PROMPT, SESSION_ID, working directory). The NPM installation script downloads pre-built binaries from GitHub releases via HTTPS, introducing a standard supply chain risk. No `eval` or `shell: true` patterns were found in the core logic. Hardcoded secrets or direct network attack surface are absent as it uses stdio transport. Overall, robust for its intended purpose.
Updated: 2025-12-05GitHub→
0
0
Medium Cost
MCP-Mirror icon
Sec9

Provides a Model Context Protocol server to enable AI assistants to interact with MongoDB databases and MongoDB Atlas resources through natural language.

Setup Requirements

  • ⚠️Requires Node.js v20 or later.
  • ⚠️Requires either a MongoDB connection string or Atlas API credentials to start the server.
  • ⚠️Requires Atlas API Service Account with appropriate permissions for MongoDB Atlas tools.
Verified SafeView Analysis
The server uses standard practices for credential management (environment variables, token-based authentication for Atlas). Sensitive information in logs is redacted using 'mongodb-redact'. Tool arguments are strongly typed with Zod schemas, mitigating injection risks. Secure password generation uses 'crypto.randomBytes'. No explicit 'eval' or direct shell execution of arbitrary user input was found. The use of EJSON for BSON type handling is safe.
Updated: 2025-12-15GitHub→
0
0
Medium Cost
poly-mcp icon
Sec9

Integrates GitLab with AI assistants like Cursor and ChatGPT for managing merge requests, CI/CD pipelines, and Architecture Decision Records (ADRs) via the Model Context Protocol (MCP).

Setup Requirements

  • ⚠️Requires a GitLab Private Access Token (glpat-) with appropriate permissions (e.g., `read_api`, `read_repository`).
  • ⚠️Python 3.8 or higher is required.
  • ⚠️For destructive operations (e.g., merging, deploying, retrying jobs/pipelines), `SAFE_MODE=false` and `DRY_RUN=false` must be explicitly set in the `.env` file.
  • ⚠️Connecting to external AI platforms like ChatGPT requires exposing the server publicly (e.g., using ngrok).
Verified SafeView Analysis
The server demonstrates strong security practices: it loads secrets from `.env` (`GITLAB_TOKEN`), implements robust input validation and URL encoding for project IDs, includes local rate limiting, and utilizes decorators (`check_safe_mode`, `dry_run_mode`, `check_project_access`) to protect destructive operations and control project access. Production deployments are explicitly guarded by an environment variable (`ALLOW_PROD_DEPLOY`). No 'eval' or malicious code patterns were found. The documentation explicitly recommends using GitLab tokens with minimal permissions. Public exposure, for example via `ngrok` for ChatGPT integration, introduces external security considerations beyond the server's control.
Updated: 2025-12-01GitHub→
0
0
Medium Cost
cirne icon

bic-mcp

by cirne

Sec6

Query and analyze Beloved In Christ Foundation grant transaction data for searching, filtering, and aggregation via an MCP server.

Setup Requirements

  • ⚠️Requires `MCP_GUID` environment variable to be set for the HTTP endpoint to be accessible (acts as a security token in the URL path).
  • ⚠️Requires transaction data files (CSV or XLSX) to be placed in the `data/` directory or a custom `DATA_DIR`.
  • ⚠️`MCP_API_KEY` is optional but highly recommended for strong authentication; without it, access relies solely on the GUID in the URL.
Verified SafeView Analysis
The server includes API key (`MCP_API_KEY`) and GUID (`MCP_GUID`) validation, which are good practices. However, the `MCP_API_KEY` is marked as optional in the documentation, and if it is not set, the `validateApiKey` function returns `true`. This means that access to the MCP endpoint relies solely on the `MCP_GUID` in the URL path, which acts more as obfuscation than a strong authentication mechanism, making deployments without `MCP_API_KEY` insecure. File operations (reading/writing CSV/XLSX) are primarily intended for local CLI scripts and do not appear to be exposed via the web API, minimizing path traversal risks.
Updated: 2025-11-28GitHub→
PreviousPage 541 of 713Next