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
Low Cost
Nagavamsi178 icon

mcp-server

by Nagavamsi178

Sec9

A simple calculator API server providing basic arithmetic operations, complemented by a Streamlit web user interface.

Setup Requirements

  • ⚠️Requires Python 3.x
  • ⚠️Dependencies include fastapi, uvicorn, pydantic, streamlit, and requests
  • ⚠️Two separate processes need to be run concurrently (FastAPI server and Streamlit UI)
Verified SafeView Analysis
The application runs locally on `127.0.0.1:8000` by default, which is generally safe. Input validation for division by zero is implemented. No 'eval' or other highly dangerous patterns are present. If deployed publicly, standard API security practices (e.g., authentication, rate limiting, more robust input sanitization) would be necessary, but this is beyond the scope of a simple calculator service.
Updated: 2025-11-25GitHub→
0
0
Medium Cost
appsmartzaiml icon

my-mcp-server

by appsmartzaiml

Sec8

Provides a Model Context Protocol (MCP) interface for AI models to search and discover live radio stations and podcasts from the Radio FM service.

Setup Requirements

  • ⚠️Requires Node.js runtime and TypeScript compiler (tsc) for building.
  • ⚠️The external Radio FM API base URL (`RADIOFM_API_BASE`) is hardcoded in `src/index.ts`.
Verified SafeView Analysis
The server uses `cors()` with `Access-Control-Allow-Origin: *` which is common for public APIs and MCP servers, but should be noted. The `RADIOFM_API_BASE` is hardcoded within the source file; changes to this external API URL would require code modification and redeployment. No usage of `eval` or other obvious malicious patterns was found. Input validation for the `query` parameter is present, and `axios` handles parameter encoding, mitigating direct injection risks in the external API call. The output formatting is plain text, preventing XSS in the response content.
Updated: 2026-01-19GitHub→
0
0
Low Cost
serge2 icon

emcp

by serge2

Sec6

Provides a lightweight framework for building MCP (Model Context Protocol) services over HTTP/JSON-RPC in Erlang.

Setup Requirements

  • ⚠️Requires Erlang/OTP environment to be installed.
  • ⚠️Requires `rebar3` build tool for dependency management and compilation.
  • ⚠️Requires secure management of API keys by the implementing application as demonstrated in the configuration example.
Verified SafeView Analysis
The framework allows application developers to define custom tool/resource handler functions. The provided example for `resources_read` demonstrates the use of `os:cmd/1` to execute shell commands. This capability, while flexible, introduces a risk of command injection if user-supplied input is passed unsanitized to `os:cmd/1` by the application developer. Secure management of API keys by the implementing application is also crucial.
Updated: 2026-01-18GitHub→
0
0
Medium Cost
dogkeeper886 icon

ruckus1-mcp

by dogkeeper886

Sec9

Facilitates automated management and querying of RUCKUS One network infrastructure (venues, APs, WiFi networks, security profiles, roles) by AI assistants via the Model Context Protocol.

Setup Requirements

  • ⚠️Requires valid RUCKUS One API credentials (Tenant ID, Client ID, Client Secret) configured as environment variables.
  • ⚠️An MCP client (e.g., Claude Desktop/CLI, Cline) is necessary to interact with the server, as it does not expose HTTP endpoints.
  • ⚠️Many 'create', 'update', and 'delete' operations are asynchronous, requiring polling with configurable retries, which means responses for these operations will include activity IDs and status details, not immediate final results.
Verified SafeView Analysis
The server uses environment variables (`RUCKUS_TENANT_ID`, `RUCKUS_CLIENT_ID`, `RUCKUS_CLIENT_SECRET`) for API credentials, which is a good practice to avoid hardcoding secrets. It communicates via standard I/O (stdio) as an MCP server, eliminating direct network exposure from the server itself. Authentication tokens are cached with an expiration buffer to minimize re-authentication and potential misuse. Error handling provides structured API responses, which helps in debugging without exposing raw internal stack traces. Input validation is performed through JSON schemas for tool arguments, and some utility validation functions exist, though their comprehensive application to all incoming data isn't fully detailed in the truncated code. There are no apparent uses of `eval` or other highly dangerous functions. Overall, the design prioritizes secure credential management and minimal network footprint for the MCP server.
Updated: 2026-01-19GitHub→
0
0
Medium Cost
tomo8812 icon
Sec7

Provides semantic PDF reading capabilities by converting PDF content to structured Markdown and extracting metadata, optimized for LLM consumption.

Setup Requirements

  • ⚠️Requires Docker for the recommended deployment, which involves mounting a volume for PDF access.
  • ⚠️The Docling library may download models on its first run, requiring an internet connection for initial setup.
  • ⚠️Requires Python 3.10 or newer.
Verified SafeView Analysis
The `read_pdf_structure` and `get_pdf_metadata` tools take a `path` argument, allowing the server to access files on the host filesystem (or within its Docker container). While the recommended Docker setup mitigates this risk by explicitly mounting a PDF-containing directory to `/data`, running the server directly via `pip` or `uv` without proper sandboxing could allow a malicious client to trigger arbitrary file reads by the server process. There are no direct `eval` calls, hardcoded secrets, or obvious malicious network patterns. The `HF_HUB_DISABLE_SYMLINKS` environment variable is explicitly set as a security measure.
Updated: 2025-11-26GitHub→
0
0
Low Cost
nazihkhelifa icon

mcp-server-templates

by nazihkhelifa

Sec1

Unable to determine the specific use case as the source code for the `mcp-server-templates` repository was not provided for analysis.

Setup Requirements

  • ⚠️Source code not provided, making it impossible to identify specific setup requirements or friction points.
Review RequiredView Analysis
The source code for the `mcp-server-templates` repository was not provided for analysis. Without access to the code, it is impossible to audit for security risks such as `eval` usage, obfuscation, network vulnerabilities, hardcoded secrets, or malicious patterns. Therefore, a score of 1 is assigned, and it is considered critically unsafe to run without a thorough code review. Any server where the source code cannot be inspected for security is inherently risky.
Updated: 2025-11-24GitHub→
0
0
Medium Cost
Aaronxx icon
Sec1

Provides a backend service for integrating applications with the WeChat platform.

Setup Requirements

  • ⚠️Requires WeChat Developer account and application credentials (AppID, AppSecret, Token).
  • ⚠️Specific runtime environment (e.g., Node.js, Python, Java) needs to be installed, but it is not specified in the provided information.
Review RequiredView Analysis
Cannot perform a comprehensive security audit as no source code was provided for analysis. Therefore, risks like 'eval', hardcoded secrets, or network vulnerabilities could exist unbeknownst. The low score reflects the inability to verify safety due to lack of visibility into the code.
Updated: 2025-11-19GitHub→
0
0
Low Cost
samaye-pro icon

mcp-server

by samaye-pro

Sec4

Demonstrates a Model Context Protocol (MCP) WebSocket server in Go for bidirectional JSON messaging and executing predefined tools.

Setup Requirements

  • ⚠️Requires Go 1.24.4+ runtime
Verified SafeView Analysis
The server's `websocket.Upgrader` is configured with `CheckOrigin: func(r *http.Request) bool { return true }`, which disables origin checking. This makes the WebSocket server vulnerable to Cross-Site WebSocket Hijacking (CSWSH) attacks if deployed in a production environment without proper remediation. While acceptable for a local demo, it's a critical security flaw for anything beyond that. There are no obvious hardcoded secrets, 'eval' equivalents, or other direct code execution vulnerabilities.
Updated: 2025-11-27GitHub→
0
0
Medium Cost
moxus icon
Sec9

Provide AI agents access to Intervals.icu data for fitness tracking, activity management, and calendar events.

Setup Requirements

  • ⚠️Requires Intervals.icu API Key (obtained from Intervals.icu website)
  • ⚠️Requires Intervals.icu Athlete ID (obtained from Intervals.icu website)
  • ⚠️Requires a Node.js runtime environment (version 18+ is recommended by dependencies)
Verified SafeView Analysis
The server securely retrieves API credentials (key and athlete ID) from environment variables, preventing hardcoding. Input and output data are rigorously validated using Zod schemas, enhancing robustness and protecting against malformed requests. API interactions are handled via Axios with basic authentication, a standard and secure method for this type of API. No 'eval' or direct arbitrary code execution vectors were identified, making the core logic safe. The reliance on `cross-spawn` (indirectly via `@modelcontextprotocol/sdk`) is for process management and not for executing user-controlled input.
Updated: 2025-12-14GitHub→
0
0
Medium Cost
Sunalamye icon

mcp-claude-node

by Sunalamye

Sec9

Acts as an MCP (Model Context Protocol) server wrapping the Claude Code CLI, enabling parallel and robust execution of Claude AI requests for code generation and editing tasks.

Setup Requirements

  • ⚠️Requires `claude` CLI to be installed and configured on the system.
  • ⚠️Requires Node.js version 18.0.0 or higher.
  • ⚠️Assumes the `claude` CLI is configured with a valid Anthropic API key for AI requests.
Verified SafeView Analysis
The server uses `node:child_process.spawn` to execute the `claude` CLI. Arguments passed to `spawn` are constructed from validated tool parameters, including specific disallowed tools to prevent certain commands. Communication is via `stdio`, limiting network exposure. The server also implements recursion depth limits (`MCP_CLAUDE_DEPTH`) to prevent infinite subprocess spawning. The `--dangerously-skip-permissions` flag is passed to the underlying `claude` CLI; while it implies broader AI permissions, it's a feature of the CLI being wrapped, not a vulnerability introduced by this server's implementation. No direct `eval` or hardcoded secrets are present.
Updated: 2025-11-30GitHub→
0
0
Medium Cost
Slaycaster icon
Sec9

Extracts key frames from Loom videos or local video files to assist with debugging and analysis in Claude Code.

Setup Requirements

  • ⚠️Docker required for deployment
  • ⚠️Requires Python 3.11 or newer
  • ⚠️Requires network access to download videos from URLs
  • ⚠️Requires host directories `~/loom-videos` (read-only input) and `~/loom-frames` (read-write output) to be available
Verified SafeView Analysis
The server uses `subprocess.run` to execute external tools (`yt-dlp` and `ffmpeg`). Commands are passed as lists of arguments, which mitigates common shell injection vulnerabilities. Network access is required for downloading videos from URLs, which is inherent to the tool's functionality. Local video input directory (`~/loom-videos`) is mounted as read-only, enhancing security. The output directory (`~/loom-frames`) is read-write, as expected.
Updated: 2025-12-15GitHub→
0
0
Medium Cost
mako10k icon
Sec7

Dynamically generates context-relevant instructions for LLMs to prevent attention dispersion and maintain focus during development.

Setup Requirements

  • ⚠️Node.js 18+ required.
  • ⚠️Requires a compatible MCP client (e.g., Claude Desktop, VS Code with GitHub Copilot).
  • ⚠️Existing monolithic `.github/copilot-instructions.md` will trigger 'Restricted Mode' and require an interactive `onboarding` migration process.
Verified SafeView Analysis
The server performs extensive file system operations (e.g., overwriting `.github/copilot-instructions.md`) and executes Git commands via `child_process.exec`. While current Git command arguments appear safely constructed, `exec` is a powerful primitive. Critical actions like migration and rollback require explicit user elicitation/approval. A file-based locking mechanism (`.copilot-state/.lock`) is in place to prevent concurrent modifications, enhancing integrity.
Updated: 2025-12-03GitHub→
PreviousPage 415 of 760Next