Shared

You run an MCP (Multi-Cloud Platform) server that exposes a library of reusable agent instructions and scripts to a compatible AI model, enabling context-aware automation across clouds. This server is designed to be bootstrapped and run locally, and then consumed by an MCP client to perform standardized tasks.

How to use

To start using your MCP server, first boot the server and then run it locally. The server is exposed via a FastAPI interface, which your MCP client can query to retrieve standardized agent instructions and utility scripts.

Once the server is running, configure your MCP client to point at the server instance. For example, if your server is reachable at http://<ip-address>:8080, add it to your client’s MCP server list under a suitable name. This enables the client to request the centralized library of agents and scripts as needed.

Example client configuration (conceptual, using a JSON-like structure): the server is referenced as shared_agents. If you follow the provided configuration snippet, your client will route requests to the MCP server for agent instructions and scripts.

How to install

Prerequisites: ensure you have Task installed to run the provided bootstrap and run commands, and a Python 3.x runtime for any server-side dependencies.

Steps to bootstrap and start the server locally:

# Bootstrap the project environment
task bootstrap

# Run the server locally
task run

Configuration and usage notes

Your MCP client should reference the server by name and URL so it can fetch the agent instructions and scripts from the central library. If you are using a client configuration similar to the example below, the client will connect to the MCP server at the specified address.

Example client configuration (as shown in the system): you configure an HTTP MCP server with a URL placeholder. The client will replace the placeholder with your actual server address at runtime.