mcpez

# MCPez - Microservices Command Proxy Management Platform [](https://opensource.org/licenses/MIT) MCPez is a web-based management platform designed to simplify the definition, configuration, management, and monitoring of backend services such as AI models, local scripts, or remote APIs. It exposes these services through standardized proxy interfaces (SSE or STDIO), making them easy to integrate and use by other applications, especially AI agents that require tool invocation. ## Project Significance The current microservices command proxy (MCP) ecosystem and AI Agent application development face several challenges: > * **Isolated MCPs have limited value**: Individual MCP services often have single functions and struggle to generate significant value independently. Users need to combine multiple MCPs to build meaningful applications, which often requires complex orchestration and integration work. > * **Complex Agent-to-Agent (A2A) interactions**: Existing A2A communication solutions are often overly complicated and lack unified standards, making it difficult to build multi-agent systems that can work together. > * **Closed ecosystems**: Many platforms lock users into specific, non-open MCP combinations, limiting flexibility and scalability. Additionally, many features of the MCP itself cannot be modified, which is counterproductive to user experience. > * **Inconsistent MCP quality**: Existing MCP platforms are filled with low-quality services, many of which are AI-generated and inadequately tested. Observations indicate that about 80% of services are repetitive, 15% have functional issues or are unavailable, and the remaining 5% often lack practical application value. This significantly increases the difficulty for developers to filter and integrate reliable tools. > _**Todo**: In the next version, there will be an MCP to assist in building the MCP server system: it is currently in testing, and the results are quite interesting._ > * **Difficult client integration and management**: Developers need to write different client adaptation code for different services and manage their configurations, deployments, and operational statuses separately, which is time-consuming and labor-intensive. Clearly, in the area of LLM using tools, agents like Lang* are increasingly moving away from outdated RAG, leading to higher token consumption, while Autogen and MCP protocols are in a state of inefficient competition, with Google also experimenting with various concepts. Pydantic is vigorously pushing its "format validation" towards standardization. This battlefield looks quite chaotic, and even Zelensky would hesitate to perform here, so don't expect a unification anytime soon. Therefore, MCPez aims to overcome these challenges by providing a more open, reliable, and manageable solution through the following means: > * **Unified management and standardized interfaces**: Encapsulating different backend services (whether local scripts, remote APIs, or other MCPs) into standardized SSE or STDIO interfaces. Providing a centralized web UI for creation, configuration, monitoring, and management, reducing management complexity. > * **Resolving client issues**: Applications (such as AI Agents) only need to interact with backend services through the stable proxy addresses provided by MCPez, without worrying about the specific implementations and deployment details of the underlying services, simplifying client integration. > * **Breaking MCP silos**: By providing an easy-to-use platform to combine and manage multiple services, originally isolated and low-value MCPs can be more easily integrated into more complex application workflows, thus generating greater value. > * **Dockerized deployment**: Providing a Dockerfile to package the entire management platform and its dependencies (such as Nginx, Python environment) for one-click deployment. This not only simplifies environment configuration but also ensures consistency in the service runtime environment, allowing users to quickly set up and run their own MCP service collections locally or on servers. > * **Improving service quality and availability**: Although MCPez does not directly write MCP services, it provides a framework that encourages users to integrate their validated, high-quality services and promotes the reuse of quality services through features like template sharing, gradually improving ecosystem quality. With MCPez, developers can focus more on building core business logic and agent capabilities rather than getting bogged down in tedious service integration and management details. ## Key Features * **Web User Interface**: An intuitive interface for managing applications and services. * **Application/Service Definition**: Supports the creation and configuration of "applications," each of which can contain multiple backend service configurations. * **Multiple Service Types**: * **SSE**: Proxies remote HTTP SSE services, supporting configuration of Base URL and Headers. * **STDIO**: Proxies local command-line processes, supporting configuration of execution commands, parameters, and environment variables. * **Configuration Management**: * Supports exporting application configurations as JSON files. * Supports importing application configurations from JSON files. * Supports saving commonly used service configurations as "tool templates" for easy reuse. * **Service Status Management**: * Lists all defined "applications" on the homepage. * Starts/stops STDIO-based services. * Views detailed status information of running services (ID, address, status, detailed logs/information). * **AI Playground**: (located at `chat.html`) Provides a chat interface where AI models (such as OpenAI, Gemini, etc.) can be configured, and services defined in MCPez can be used as Tool/Function Calling backends, facilitating testing and debugging of AI interactions with tools. * **Docker Support**: Provides a Dockerfile for easy containerized deployment. ## Future Directions * **AI-driven tool descriptions and function understanding**: Plans to introduce features that allow large language models (LLMs) to automatically generate or optimize descriptions of tools (MCP services) and their internal functions (if applicable) based on service configurations and context. This will greatly enhance AI Agents' understanding of tool functionalities and effectively address the current issue of low hit rates when selecting and invoking MCP functions. * **Shareable toolsets**: Users can export carefully configured and validated applications (containing a set of related MCP services) as configuration packages. These configuration packages can be shared within the community or team, allowing other users to easily import and reuse these toolsets, accelerating the development process. * **Local-first and security**: The toolset sharing mechanism will maintain the core advantages of localized deployment. After importing configurations, services will still run locally, avoiding the security risks associated with uploading sensitive information (such as API Keys) to third-party sharing platforms. This ensures the security of data and credentials while promoting the circulation of trusted tools. **Disclaimer**: Indirectly helps you reduce taxes by 104%. ## Quick Start ### Prerequisites * [Docker](https://www.docker.com/) and [Docker Compose](https://docs.docker.com/compose/) (recommended) ### Running with Docker 1. **Build the Docker image**: ```bash docker build -t MCPez . ``` 2. **Run the Docker container**: ```bash docker run -d -p 8088:80 --name MCPez-instance -v MCPez_data:/data MCPez ``` * `-d`: Run in detached mode. * `-p 8088:80`: Maps port 8088 on the host to port 80 in the container (Nginx default port). You can change the host port `8088` as needed. * `--name MCPez-instance`: Names the container. * `-v MCPez_data:/data`: Creates a Docker volume `MCPez_data` to persistently store SQLite database and service configuration data. ### Accessing the Web UI Open `http://localhost:8088` (or the host port you specified) in your browser.      ### Usage Instructions 1. **Service Management (`index.html`)**: * The main interface displays a list of all created applications and their statuses. * Applications can be searched. * For STDIO type services, you can click the "Start" or "Stop" button. * Click "Edit Service" to navigate to the editing page. * Click "View Service Status" to see detailed information about running services. * Click "New Service" to navigate to the application editing page. 2. **Application Editing (`edit.html`)**: * **Create New Application**: Access `edit.html` directly. * **Edit Existing Application**: Click "Edit Service" from the homepage or access via `edit.html?id=<app_id>`. * **Application Information**: Set the application's name and description. * **Configuration Management**: * **Import/Export**: Use JSON files to import or export the entire application configuration (including name, description, and all server configurations). * **Using Templates**: Select and add service configurations from the pre-stored tool template library. * **Server Configuration**: * Click the "MCP" button to add new server configurations (SSE or STDIO). * Fill in the server name, type (SSE/STDIO), description, and corresponding configurations (URL/Headers or Command/Args/Env). * Edit or delete existing server configurations. * Configured servers can be saved as "tool templates" for future use. * **JSON Preview**: Displays the complete JSON configuration of the current application in real-time. * **Save Configuration**: Click the "Save Configuration" button in the upper right corner to save changes. 3. **AI Playground (`chat.html`)**: * Click the gear icon in the upper right corner to configure settings, including API Key, Base URL, model name, etc., for the AI model provider. * Configure the MCP server address (usually `http://localhost:8088/mcp/<app_id>/sse`, where `<app_id>` is the ID/name of the application configured in `edit.html`). * Once configured, you can chat with the AI in the chat interface. If the AI model supports Tool/Function Calling and the configured MCP application has corresponding services, the AI will be able to invoke these services. ## License This project is licensed under the [MIT License](LICENSE). ``` MIT License Copyright (c) [Year] [Copyright Holder] Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ```