Discorevy local MCP Servers. Specification.

What is Discorevy local MCP Servers. Specification.?

It is a specification that defines a standard way to list and configure MCP (Model Context Protocol) servers on a local machine. It uses markdown files stored in the ~/.mcp folder, which an LLM‑powered client interprets to discover and set up MCP servers.

How to use Discorevy local MCP Servers. Specification.?

Create a markdown file per MCP server inside ~/.mcp (or %USER_HOME% on Windows). The file describes the server’s name, ID, version, URL, authentication, capabilities, regions, health check endpoint, and metadata. The MCP client uses an LLM (e.g., Claude, ChatGPT) to extract the necessary information from these files and decides whether and how to use the server. The client must periodically refresh the files to discover new or updated servers.

Key features of Discorevy local MCP Servers. Specification.

• Standard, local‑only file‑based MCP server registration.
• Markdown format understood by LLMs (Claude, ChatGPT, etc.).
• Client‑driven discovery via periodic disk scans.
• Includes fields for authentication, health checks, and metadata.
• No security implications handled – left to MCP clients.

Use cases of Discorevy local MCP Servers. Specification.

• Registering production or development MCP servers on a developer’s machine.
• Enabling AI agents to discover and invoke local tools.
• Standardising how different IDEs (e.g., IntelliJ IDEA) and chat clients (Claude, ChatGPT) find MCP servers.
• Managing multiple environments (e.g., production, staging) with distinct credentials and endpoints.

FAQ from Discorevy local MCP Servers. Specification.

What is the purpose of the specification?

To provide a standard way to list and configure MCP Servers on a local machine, currently missing from many MCP clients.

Where are MCP server definitions stored?

In the ~/.mcp folder (or %USER_HOME% on Windows). Each server gets its own markdown file.

Which tools support this specification?

As of v1.0.1, it is unsupported by default; the author invites pull requests from tools like IntelliJ IDEA, Claude, ChatGPT, Cursor, Windsurf, and Warp.

How does a client discover and use a server?

The client regularly reads all markdown files under ~/.mcp and uses an LLM to extract configuration (authentication, endpoints, capabilities, etc.). The LLM and client decide whether to use the server and how to handle credentials.

Does the specification handle security?

No. Security implications for MCP servers remain the responsibility of MCP clients.