MCPO-Simple serves as a Python-based MCP (Model Context Protocol) server written to run MCP tools inside isolated, secure environments. It exposes access to MCP tools over SSE, Swagger REST, and an OpenAPI.json (compatible with OpenWebUI), and runs each tool in its own thread for isolation. The server is designed to be started via a tool loader (uvx) or similar launcher, enabling quick loading of MCP tools by package name. You can interact with the server's endpoints to list available MCP tools, start a tool, and stream or query responses through the MCP interface. The included environment-based admin hack lets you simulate a Bearer-admin authorization flow for testing and quick changes without a full auth backend.

To use, start the server with your preferred launcher (uvx in this configuration) and load MCP tools by their package names. Once running, you can browse the Swagger UI or OpenAPI.json to see the available MCP tools and their capabilities, then invoke a tool and receive execution results or streaming outputs via the provided endpoints.

Prerequisites:

• Python 3.11+ (or a compatible Python environment)
• Access to a shell/terminal
• Optional: Docker if you prefer containerized usage

Installation steps:

1. Clone the repository (or download the distribution): git clone https://github.com/getsimpletool/mcpo-simple-server.git cd mcpo-simple-server

2. Create a Python virtual environment (recommended): python -m venv venv source venv/bin/activate # on Windows: venv\Scripts\activate

3. Install Python dependencies: pip install -r requirements.txt

4. Install uvx (the tool loader used to load MCP tools): pip install uvx

5. Install or prepare MCP tools you want to load (example): uvx install <tool-package-name>

   e.g., uvx install getsimpletool-sample-tool

6. Run the MCP server using the configured launcher (uvx in this setup): uvx run getsimpletool-mcpo-simple-server

7. Verify the server is running by visiting: http://localhost:8000/docs (Swagger UI) http://localhost:8000/openapi.json

Notes:

• If you prefer Docker, you can containerize this server by creating a Dockerfile that installs dependencies and runs the uvx-based launcher.
• Adjust environment variables as needed (see the environment section).

• The server uses a JSON-based configuration for portability and easy editing. Keep your mcpServers map up to date with the tools you want to expose.
• The Bearer admin simulation (ENV var) is intended for testing and quick changes without implementing a full auth system.
• If you encounter tool loading issues, ensure the tool package names match exactly what uvx expects and that network access to tool registries is available.
• Use the Swagger/OpenAPI endpoints to understand the available MCP tools, their inputs, and output formats. This helps in building client code that consumes the MCP API.
• For debugging, enable verbose logging by setting MCP_SERVER_LOG_LEVEL to debug or trace.