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
blockrunner icon

landcomp-mcp

by blockrunner

Sec6

Acts as an MCP (Model Context Protocol) server to integrate with the LandComp application, providing tools for plant catalog search, landscape composition creation, and visualization, supporting both local (stdio) and Docker (SSE) modes.

Setup Requirements

  • ⚠️Requires Docker and a shared Docker network (`landcomp_network`) for production/SSE mode.
  • ⚠️Depends on an external LandComp API, specified by `LANDCOMP_API_URL`.
  • ⚠️An `MCP_API_KEY` is required for authentication with the LandComp API.
Verified SafeView Analysis
The server uses permissive CORS headers (`Access-Control-Allow-Origin: *`) in SSE mode, which is generally not recommended for public-facing APIs without stricter origin controls. Debug logging can expose sensitive request/response data, including API keys, if not configured appropriately for production. The example API key `test-local-api-key-12345` is publicly known from configuration examples, which is a weak default for local testing.
Updated: 2025-11-19GitHub→
0
0
High Cost
AnotherRegularDude icon

ollama-web-search-mcp

by AnotherRegularDude

Sec9

Provides web search and web content fetching capabilities to AI assistants (LLMs) via the Model Context Protocol (MCP), using the Ollama web search API.

Setup Requirements

  • ⚠️Requires Ruby 3.4.8 and Bundler.
  • ⚠️Requires an `OLLAMA_API_KEY` environment variable, which may be a paid service.
Verified SafeView Analysis
The server employs strong typing (Dry::Struct, Dry::Types) and validation for all incoming parameters, reducing risks of basic injection attacks. External API calls to 'ollama.com' are made securely over HTTPS, with parameters serialized to JSON. The Ollama API key is correctly sourced from environment variables, preventing hardcoding. The layered architecture and explicit service object pattern (Resol::Service) promote robust error handling and separation of concerns, limiting internal vulnerabilities. While it proxies user input to an external API, the application's internal handling of this input is designed for safety.
Updated: 2026-01-06GitHub→
0
0
High Cost
Tommertom icon

gemini-mcp

by Tommertom

Sec9

An AI-powered Model Context Protocol (MCP) server for Google Gemini multimodal media operations, specifically designed to empower coding agents and AI development workflows with image generation, analysis, and manipulation instructions.

Setup Requirements

  • ⚠️Requires a Google Gemini API Key, which must be obtained from Google AI Studio and configured in a .env file.
  • ⚠️The project must be built using 'npm run build' before it can be run as a standalone executable.
  • ⚠️Output files are saved to '/tmp/gemini_mcp' by default, which may require awareness for users on Windows or those expecting a different output location.
Verified SafeView Analysis
The server correctly handles the GEMINI_API_KEY via environment variables, preventing hardcoding. File system operations (read/write) are performed within designated output directories (e.g., /tmp/gemini_mcp) or on client-provided paths for analysis. While client-provided file paths could theoretically lead to issues if an agent points to sensitive system files, the server's direct file operations (fs.readFile) are generally safer than executing arbitrary commands. No 'eval' or direct arbitrary command execution from user input is observed. The testing script uses 'child_process.spawn' to start the server, which is appropriate for testing and not a security risk for the server itself.
Updated: 2025-11-24GitHub→
0
0
High Cost
jayyprajapati icon

Chatbot-the-MCP-way

by jayyprajapati

Sec3

Interactive command-line chat application enabling AI model interaction with document retrieval and extensible command execution via the Model Control Protocol (MCP).

Setup Requirements

  • ⚠️Requires Ollama to be installed and running locally.
  • ⚠️Requires a Llama model (e.g., llama2) to be pulled in Ollama.
  • ⚠️Requires manual configuration of LLAMA_MODEL and OLLAMA_BASE_URL in a .env file.
Review RequiredView Analysis
The `MCPClient` in `mcp_client.py` is designed to execute arbitrary commands (`command` and `args`). While the `doc_client` in `main.py` uses fixed, safe commands for `mcp_server.py`, the application also iterates through `sys.argv[1:]` to launch additional `MCPClient` instances with `uv run server_script`. This allows any script passed as a command-line argument to `main.py` to be executed. This poses a significant command injection risk if `main.py` is ever invoked with untrusted arguments, as no input validation or whitelisting is evident for these additional server scripts. Document content editing via `edit_document` also presents a minor data integrity risk if the LLM is unconstrained.
Updated: 2025-12-03GitHub→
0
0
Medium Cost
EugeneAnt icon
Sec8

Enables AI assistants to directly parse, read, and modify SDLXLIFF translation files for localization tasks.

Setup Requirements

  • ⚠️Designed for Model Context Protocol (MCP) environments (e.g., Claude Desktop/Cowork)
  • ⚠️File path resolution tailored for Claude Cowork sandbox, may behave unexpectedly in other environments.
Verified SafeView Analysis
The server primarily interacts with the local file system via standard input/output, mitigating typical network-related risks. It uses `lxml` for XML parsing, which is generally robust. The `resolve_file_path` function attempts to translate sandbox paths to host paths by searching common user directories (Documents, Downloads, Desktop) and uses `Path.resolve()` to canonicalize paths, which helps prevent directory traversal. XPath values are handled with some basic escaping and iteration for complex IDs, reducing XPath injection risk. The main security consideration is the scope of file access granted by the execution environment (e.g., Claude Cowork sandbox) and the potential for a malicious file_path input to influence file discovery within the defined search roots. No hardcoded secrets or 'eval' statements were found.
Updated: 2026-01-19GitHub→
0
0
Low Cost
dmarsters icon
Sec9

An MCP server that translates jazz improvisation structure, particularly centered on Thelonious Monk's approach, into visual sensory parameters for enhancing AI image generation prompts.

Setup Requirements

  • ⚠️Requires Python 3.10 or higher.
  • ⚠️Requires `fastmcp` and `pyyaml` Python packages.
Verified SafeView Analysis
Uses `yaml.safe_load` for configuration, mitigating YAML injection risks. No `eval`, `exec`, `subprocess`, or hardcoded credentials found. Input parameters are type-hinted and range-checked.
Updated: 2025-11-29GitHub→
0
0
Medium Cost
pokidyshev icon

oura-mcp

by pokidyshev

Sec8

The server connects Oura Ring health data to AI assistants via the Model Context Protocol (MCP) for natural language querying and analysis.

Setup Requirements

  • ⚠️Requires obtaining a Personal Access Token from Oura's developer portal for personal use.
  • ⚠️For production OAuth2 deployment, requires setting OURA_CLIENT_ID, OURA_CLIENT_SECRET, DEPLOYED_URL, and a securely generated JWT_SIGNING_KEY as environment variables.
  • ⚠️The internal OuraClient's OAuth2 token refresh mechanism is marked as 'NOT TESTED YET', which could be a concern if relying solely on that specific client-side refresh path rather than FastMCP's OAuthProxy.
Verified SafeView Analysis
Secrets (API keys, tokens, JWT signing key) are appropriately loaded from environment variables, which is good practice. A local '.oura_tokens.json' file is used for PAT persistence but explicitly marked for 'LOCAL/DEV/TESTING ONLY'. The primary OAuth2 token management is handled by FastMCP's OAuthProxy, which should abstract away most security concerns for production. However, a warning '⚠️ WARNING: OAuth2 token refresh is NOT TESTED YET!' within the OuraClient class suggests potential fragility or an incomplete client-side token refresh implementation in that specific path, though the broader FastMCP OAuth flow is expected to be robust. The JWT_SIGNING_KEY is critical for OAuth2 and must remain stable.
Updated: 2025-11-23GitHub→
0
0
High Cost
Daulet151 icon

hj-mcp-server

by Daulet151

Sec6

AI-powered natural language SQL query generation and data analysis for the Hero's Journey data warehouse, integrated with Slack and Model Context Protocol (MCP) clients.

Setup Requirements

  • ⚠️Requires OpenAI API Key (paid service, access to gpt-4o model recommended)
  • ⚠️Requires access to a PostgreSQL database (Hero's Journey DWH)
  • ⚠️Slack integration requires a configured Slack bot token, signing secret, and specific permissions (channels:history, chat:write, files:write)
  • ⚠️Deployment to production environments for the Slack bot typically requires HTTPS/SSL for the webhook URL.
Verified SafeView Analysis
The `execute_sql` MCP tool directly executes user-provided SQL after a basic `SELECT` prefix check, which is susceptible to SQL injection if the database user has broader permissions than strictly `SELECT`. The system logs all generated and executed SQL queries, which could inadvertently capture Personally Identifiable Information (PII) if sensitive data is used in user prompts that get reflected in the SQL. While the documentation recommends read-only database access and marks PII fields, enforcement relies on external configuration and careful usage. No evident use of `eval` or deliberate obfuscation.
Updated: 2026-01-16GitHub→
0
0
Medium Cost
timesketch icon
Sec7

Provides an API server with a set of tools for programmatic interaction and analysis of data within a Timesketch digital forensics platform.

Setup Requirements

  • ⚠️Requires Docker and Docker Compose for easy setup.
  • ⚠️Requires an existing and accessible Timesketch instance with configured credentials.
  • ⚠️Requires Python 3.11 or newer if running directly without Docker.
Verified SafeView Analysis
The server uses environment variables for Timesketch credentials, which is good practice. Input sanitization for `search_timesketch_events_substrings` handles Lucene reserved characters for non-regex searches. However, tools like `search_timesketch_events_advanced` allow direct Lucene/OpenSearch query strings, meaning the security against query injection or resource exhaustion largely depends on the robustness of the underlying `timesketch-api-client` and the Timesketch backend itself. The server binds to `0.0.0.0` by default when run via `uv`, making it accessible externally, requiring proper network isolation in production.
Updated: 2026-01-12GitHub→
0
0
Medium Cost
niroe5tar64 icon

niro-mcp-servers

by niroe5tar64

Sec9

Retrieves Confluence page content and page tree structures, converting them into a clean Markdown format optimized for AI agent consumption and knowledge management.

Setup Requirements

  • ⚠️Docker and Docker Compose are required to run the server.
  • ⚠️Requires Confluence API credentials (CONFLUENCE_USERNAME and either CONFLUENCE_PASSWORD or CONFLUENCE_API_TOKEN) for a Confluence instance.
  • ⚠️The Confluence user/API token must have appropriate read permissions to access the desired spaces and pages.
Verified SafeView Analysis
The server is designed for secure execution within Docker containers, featuring a read-only filesystem, no-new-privileges, and stdio-only communication with Claude Desktop. It utilizes reputable libraries for HTML parsing and Markdown conversion. The primary security concern lies in the management of Confluence API credentials, with the README explicitly recommending Docker secrets or other secure secret management systems for production environments.
Updated: 2026-01-05GitHub→
0
0
Low Cost
UnitVectorY-Labs icon

mcp-acronym-lookup

by UnitVectorY-Labs

Sec9

Provides an MCP server that resolves acronyms and initialisms from a CSV file.

Setup Requirements

  • ⚠️Requires the ACRONYM_FILE environment variable to be set to the path of a three-column CSV file.
  • ⚠️CSV file must be UTF-8 encoded with columns: `acronym`, `full form`, `description`.
  • ⚠️Installation requires a Go environment for building from source or downloading pre-compiled executables.
Verified SafeView Analysis
The server is a simple lookup service with minimal external dependencies. Input sanitization is performed on lookup keys (case-insensitive, non-alphabetic characters removed via regex). No 'eval' or similar dangerous patterns, hardcoded secrets, or obvious malicious code. Risks are primarily confined to the contents of the user-provided CSV file and the exposure of the HTTP server if not properly secured in deployment.
Updated: 2026-01-17GitHub→
0
0
Low Cost
SZ1995-sz icon

mcp_server

by SZ1995-sz

Sec9

This server serves as a quickstart example for FastMCP, demonstrating how to define and expose AI tools, resources, and prompt generators.

Setup Requirements

  • ⚠️Requires Python 3.12 or higher
  • ⚠️Relies on 'mcp[cli]' package version 1.22.0 or newer
  • ⚠️Requires the 'uv' tool for running as per the suggested command
Verified SafeView Analysis
The provided source code is a simple demo and does not contain obvious security vulnerabilities such as 'eval' or hardcoded secrets. Inputs are used in a controlled manner (integer arithmetic, f-strings for greetings). Security would primarily depend on the underlying 'mcp' framework and how it handles external interactions.
Updated: 2025-11-23GitHub→
PreviousPage 652 of 713Next