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)

32
2
Low Cost
InfraMCP icon
Sec7

Enables AI agents to interact with Atlassian Cloud services (Jira, Confluence, Jira Service Management) for work documentation, issue management, and context understanding through a unified API.

Setup Requirements

  • ⚠️Requires an Atlassian OAuth 2.0 app to be created and configured in the Atlassian Developer Console.
  • ⚠️Specific OAuth scopes (e.g., `read:jira-work`, `write:page:confluence`, `read:servicedesk-request`, `offline_access`) must be explicitly configured in the Atlassian OAuth app's permissions.
  • ⚠️The OAuth callback URL must be set exactly to `http://localhost:8080/callback` in the Atlassian OAuth app settings, which can cause port conflicts locally.
Verified SafeView Analysis
The server implements OAuth 2.0 with PKCE for strong authentication. However, several security risks were identified in the project's own `SECURITY_ASSESSMENT.md` and through code analysis: 1. **Hardcoded Localhost Callback Server**: The OAuth callback server binds to `localhost:8080` without configuration, posing risks of port conflicts and potential local hijacking. 2. **Client Secret in Environment Variables**: Storing the OAuth client secret in plain text environment variables makes it visible in process lists and logs. 3. **Credential File Storage**: Access and refresh tokens are stored in a local JSON file (`~/.atlassian_mcp_credentials.json`) with `0600` permissions, but without encryption at rest, allowing access by anyone with access to the user's account and potentially persisting indefinitely. 4. **Limited HTTP Callback Server Security**: The temporary HTTP server for OAuth callbacks has minimal security, lacking comprehensive request validation and robust timeouts, making it potentially vulnerable to manipulation or DoS during the callback window. 5. **Error Information Disclosure**: Detailed error messages could potentially leak sensitive system information in a production environment. While the code itself does not appear to contain malicious patterns like `eval` or obfuscation, the handling of sensitive credentials locally and the network configuration for the OAuth callback require careful consideration for production deployments.
Updated: 2025-12-14GitHub→
32
2
Medium Cost
boecht icon

birre

by boecht

Sec9

BiRRe is a Model Context Protocol (MCP) server that translates a BitSight subscription into LLM-friendly tools for security rating retrieval and risk management.

Setup Requirements

  • ⚠️Requires a BitSight API key (implies a paid BitSight subscription).
  • ⚠️Requires Python 3.13+ runtime environment.
  • ⚠️Requires HTTPS network access to `api.bitsighttech.com`; custom CA certificates might be needed for corporate proxies.
Verified SafeView Analysis
The project follows good security practices. API keys are required as environment variables, not hardcoded. Network communication via `httpx` enforces TLS 1.2+ by default, with explicit options for custom CA bundles or disabling verification (with warnings) for testing. Input validation is performed for CLI arguments and some API parameters. No `eval` or `exec` calls were found. File operations are standard for configuration. The `--allow-insecure-tls` flag exists for diagnostics, which is a potential risk if misused, but is clearly documented and warns the user.
Updated: 2026-01-19GitHub→
32
91
Medium Cost
mahdin75 icon

gis-mcp

by mahdin75

Sec2

A Model Context Protocol (MCP) server that provides AI agents and LLMs with comprehensive GIS capabilities, enabling geospatial analysis, data gathering, and transformations through natural language.

Setup Requirements

  • ⚠️Requires Python 3.10 or higher.
  • ⚠️Specific features (e.g., visualization, climate data) require installation of optional dependencies (e.g., `uv pip install gis-mcp[climate]`).
  • ⚠️Climate data download (`cdsapi`) requires an external Copernicus Climate Data Store (CDS) account, API key, and acceptance of license agreements, stored in a `~/.cdsapirc` file.
  • ⚠️Critical security vulnerability related to arbitrary file access if user-provided paths are not strictly validated and confined.
Review RequiredView Analysis
CRITICAL: The `resolve_path` function in `src/gis_mcp/storage_config.py` allows user-provided absolute paths for file operations (upload, download, list, and data-specific output paths). This enables arbitrary file read/write/list operations anywhere on the server's filesystem, limited only by the permissions of the running process, leading to a severe path traversal / arbitrary file operation vulnerability. Additionally, when running in HTTP mode, the server defaults to `0.0.0.0` (all interfaces), making it publicly accessible.
Updated: 2025-12-21GitHub→
32
7
Medium Cost
StacklokLabs icon

plotting-mcp

by StacklokLabs

Sec8

A MCP server that transforms CSV data into beautiful visualizations for AI assistants and chat applications.

Setup Requirements

  • ⚠️Python 3.13+ required.
  • ⚠️Requires a writable filesystem for Matplotlib/Cartopy caching (especially in containerized environments).
  • ⚠️Docker is required for ToolHive deployment.
Verified SafeView Analysis
The server uses `json.loads` on user-provided `json_kwargs`. While Python's `json` module is generally safe against arbitrary code execution, extensive or malformed inputs could potentially lead to resource exhaustion or unexpected plotting library behavior. The server listens on `0.0.0.0`, requiring external network security. No `eval` or hardcoded secrets were found. The project includes a `SECURITY.md` and a responsible disclosure process.
Updated: 2026-01-07GitHub→
32
2
High Cost
ianaleck icon
Sec9

Provides a Model Context Protocol (MCP) server for seamless integration with the Harvest time tracking API, enabling AI clients to manage time entries, projects, clients, users, expenses, invoices, estimates, and reports.

Setup Requirements

  • ⚠️Requires a Harvest account with API access (Personal Access Token and Account ID).
  • ⚠️Requires manual configuration in an MCP-compatible client (e.g., Claude Desktop) with specific JSON syntax for `command` and `env` variables.
  • ⚠️Harvest API responses, especially for list operations with high `per_page` values, can be very verbose, potentially leading to high LLM token costs when the output is consumed by an AI.
Verified SafeView Analysis
The server demonstrates strong security practices. It uses Zod for comprehensive input validation on all tool arguments, significantly reducing the risk of injection attacks or malformed requests. Sensitive API credentials (Harvest Access Token and Account ID) are strictly loaded from environment variables, preventing hardcoding. Error handling is well-structured, wrapping exceptions in MCP-compliant error responses to prevent sensitive information disclosure. Dependencies like Axios, Express, Winston, and Zod are reputable. An optional API key authentication is available for the HTTP transport, which is a good security layer, although not mandatory by default for HTTP.
Updated: 2025-12-01GitHub→
32
2
Medium Cost
QuentinCody icon

entrez-mcp-server

by QuentinCody

Sec9

Provides a comprehensive Model Context Protocol (MCP) server that enables LLMs and applications to interact with NCBI biomedical databases (E-utilities, PubChem, PMC, BLAST) using language-native SDKs and advanced SQL-based data staging for analysis.

Setup Requirements

  • ⚠️An optional NCBI_API_KEY environment variable is highly recommended to increase API rate limits from 3 to 10 requests/second (3.3x faster).
  • ⚠️Designed for deployment on Cloudflare Workers, requiring a Cloudflare account for production use.
  • ⚠️For local development and SDK testing, the server needs to be run locally via 'npm start' or 'npx wrangler dev'.
Verified SafeView Analysis
The server explicitly enforces 'SELECT' only queries for staged data, mitigating SQL injection risks. NCBI API keys are loaded securely from environment variables, not hardcoded. External API calls are directed to well-known NCBI endpoints. No 'eval' or obfuscation detected. Hardcoded default email/tool are for identification with NCBI, not sensitive.
Updated: 2025-12-04GitHub→
32
2
Low Cost
jasonkneen icon

fluent-mcp

by jasonkneen

Sec9

A fluent API for building Model Context Protocol (MCP) servers and clients to orchestrate AI workflows and agent interactions across various platforms with built-in CRUD and transport management.

Setup Requirements

  • ⚠️Requires Node.js runtime (>=18.0.0).
  • ⚠️MCP is a protocol, so understanding its concepts (tools, resources, prompts) is necessary.
  • ⚠️Default resource storage is in-memory; persistent storage requires custom integration.
  • ⚠️Using SSE transport requires an Express.js server setup to handle HTTP requests.
Verified SafeView Analysis
The core library is built on `@modelcontextprotocol/sdk` and uses Zod for schema validation in tool definitions, which is good for input sanitization. The framework allows developers to define custom tool handlers, and the overall security depends on how these handlers are implemented by the user. No direct `eval` or explicit command injection vulnerabilities were found in the library's core implementation. Console output redirection for stdio transport is a good practice for MCP server integrity. Running the server with network transports (like SSE) on `0.0.0.0` is standard but requires proper environment setup (e.g., firewall) for production.
Updated: 2025-11-26GitHub→
32
2
Medium Cost
cvenkatreddy icon

playwright-mcp

by cvenkatreddy

Sec8

Provides a guide and example client project for generating and running Playwright UI and API tests using an MCP Playwright server with LLM agents.

Setup Requirements

  • ⚠️Requires Node.js v18 or newer.
  • ⚠️Requires installation of the '@playwright/mcp' package.
  • ⚠️Requires a compatible MCP client (e.g., Claude Desktop, Windsurf, VS Code) and integration with an LLM agent like Claude Sonnet 4 for test generation.
Verified SafeView Analysis
The provided source code for the client project, including test files and utility scripts, does not contain direct security vulnerabilities like 'eval' or hardcoded secrets. It relies on `npx @playwright/mcp@latest` to start the server, an external dependency from a trusted source (Microsoft). The primary security consideration arises from the nature of LLM-generated code, as the README states that generated test suites can vary. Running unreviewed, dynamically generated tests could pose risks if the LLM produces malicious code, although the provided example run commands explicitly target specific, benign test files included in the repository.
Updated: 2026-01-09GitHub→
32
2
Low Cost
Siam-analytics icon

claude-code-lsps

by Siam-analytics

Sec9

Enhance coding experience by providing Language Server Protocol (LSP) plugins for various programming languages with features like code completion and error checking.

Setup Requirements

  • ⚠️Requires specific language runtimes and package managers (e.g., Go, Java 21+, Node.js/npm, Python/pip, Ruby/gem, .NET SDK) depending on the desired LSP.
  • ⚠️macOS users benefit from Homebrew for simplified installation of several LSPs.
Verified SafeView Analysis
The source code primarily consists of shell scripts (`.sh`) designed to check for and install various Language Servers (e.g., gopls, jdtls, pyright) using standard package managers (go install, brew, npm, pip, gem). There is no use of `eval`, obfuscation, hardcoded secrets, or direct network calls beyond what the package managers themselves perform. The scripts are straightforward and perform common development environment setup tasks. The 'risk' is inherent in any system that automatically installs software, but the commands used are standard and well-understood for their respective tools.
Updated: 2026-01-19GitHub→
32
2
Medium Cost
AD-Archer icon
Sec8

This project acts as a Model Context Protocol (MCP) bridge, enabling OpenWebUI or other MCP-capable clients (like n8n) to communicate with an internal, in-house AI platform via HTTP webhooks, providing conversation memory and OpenAI-compatible endpoints.

Setup Requirements

  • ⚠️Requires `AI_WEBHOOK_URL` to be configured, pointing to an internal AI service.
  • ⚠️Network setup often requires ports 80/443 forwarding or tunneling solutions (e.g., Tailscale, Cloudflare Tunnels) for the AI service to communicate back to the bridge.
  • ⚠️Requires Python 3.10+ if running locally, or Docker for containerized deployment.
Verified SafeView Analysis
The server uses Pydantic for configuration validation, `sqlite3` with parameterized queries to prevent SQL injection, and provides a `BearerAuthMiddleware` for API authentication. It includes retry logic for AI webhook calls. A notable feature is the `trigger_webhook` tool which can invoke arbitrary URLs if not targeting a pre-configured alias, posing a potential Server-Side Request Forgery (SSRF) risk if the bridge is exposed without strong authentication or if an attacker gains control of the MCP client. However, this is described as an intended feature for flexibility. No 'eval' or obvious obfuscation/malicious patterns were found.
Updated: 2026-01-09GitHub→
32
2
Low Cost
epicmotionSD icon

openconductor

by epicmotionSD

Sec2

Managing and orchestrating MCP (Model Context Protocol) servers for AI agent development, offering discovery, installation, and lifecycle management.

Setup Requirements

  • ⚠️Requires Node.js 18+ and PostgreSQL (or Supabase account) for the backend API.
  • ⚠️Manual database migration (`pnpm run db:migrate`) and environment variable setup are necessary for local API development.
  • ⚠️The presence of hardcoded production database credentials in various scripts poses a critical security risk upon cloning or running any script from the repository, potentially granting access to sensitive data.
Review RequiredView Analysis
Multiple shell scripts (`scripts/run-stacks-migration.sh`, `scripts/add-server.sh`, `scripts/pull-supabase-data.sh`) contain hardcoded production PostgreSQL database credentials (username, password, host, port, database) and Supabase API keys directly within the repository. The `vercel.json` (as described in `docs/archive/DEPLOYMENT_FIX_SUMMARY.md`) also embeds a full PostgreSQL connection string. The `ssl: { rejectUnauthorized: false }` setting in `api/v1/servers.ts` can introduce MITM vulnerabilities if not used in a highly controlled environment. The public API has `Access-Control-Allow-Origin: *` which is common for public APIs but less secure than restricting origins. Environment variables are not consistently used across all deployment and setup scripts, leading to potential credential exposure.
Updated: 2026-01-19GitHub→
32
2
Low Cost
andreibesleaga icon

oscp-sdk

by andreibesleaga

Sec8

Enables AI assistants to interact with EV charging infrastructure using the Open Smart Charging Protocol (OSCP) via a Model Context Protocol (MCP) server.

Setup Requirements

  • ⚠️Requires 'OSCP_API_KEY' environment variable to be set.
  • ⚠️Requires Node.js 20 LTS or later.
  • ⚠️The 'Code Mode' feature depends on an external remote execution service (Stainless API) and requires trust in its sandbox isolation.
Verified SafeView Analysis
The server implements a 'Code Mode' tool that allows AI agents to write and execute TypeScript code against the SDK. This code execution is explicitly stated to occur in an 'isolated sandbox' on a remote Stainless API service (api.stainless.com/api/ai/code-tool), without local web or filesystem access. While delegating execution to a sandboxed remote service reduces local risk, it relies heavily on the security and isolation of that external service. No direct local 'eval' or similar dangerous patterns are used for agent code within the provided server code. API keys (OSCP_API_KEY) are obtained from environment variables or client options, not hardcoded. Authentication for HTTP transport is via Bearer token or x-oscp-api-key header.
Updated: 2026-01-19GitHub→
PreviousPage 170 of 713Next