MCP Server Templates is a deployment platform for the Model Context Protocol (MCP), designed to let developers quickly run MCP servers without needing to understand infrastructure in depth. It provides Docker and Kubernetes backends, a lightweight command-line tool mcpt, and client utilities for seamless MCP integration.
Key capabilities include:
- One-click deployment: instantly deploy MCP servers using prebuilt templates, with no Docker knowledge required.
- Smart tool discovery: automatically discover and display all tools provided by the server.
- Smart caching: templates are cached for 6 hours, with automatic invalidation supported, to speed up operations.
- Comprehensive command-line interface: for deployment, management, and tool execution.
- Flexible configuration: configurable via JSON, YAML, environment variables, CLI options, or override parameters.
- A growing template library: ready-to-use templates for common use cases such as file systems, databases, APIs, and more.
Usage: Can be installed via PyPI (recommended). After deployment, the MCP server runs by default on http://localhost:8080。支持从模板启动服务器、通过单一端点进行负载均衡路由，并同时支持已部署（HTTP）和本地（stdio）传输。配置流程为：模板默认值 → configuration file → CLI options → environment variables.
Template development: You can use the generator mcpt create my-template to create a template, and define template.json, which includes the name, description, Docker image, transport method (default stdio, supporting stdio and http), and configuration schema (such as environment variable mapping for api_key and sensitive flags). The test and deployment command is mcpt deploy my-template --backend mock.
Use cases: Suitable for AI developers, data scientists, and DevOps teams when building MCP-related applications, especially in scenarios that require rapid deployment, management of MCP servers, and integration of AI capabilities.
Limitations and migration notes: The repository has been renamed to MCP Platform, and the original version is in maintenance mode, providing only security updates until 2025. The new repository is Data-Everything/MCP-Platform, the new package is pip install mcp-platform (replacing mcp-templates), and the new CLI is mcpp (replacing mcpt). Migration steps include: uninstall the old package and install the new package, replace mcpt with mcpp in commands, and update documentation bookmarks. Existing configurations, Docker images, templates, and deployment workflows remain compatible. The project is licensed under the Elastic License 2.0.
