dsh-mcp-manage

DSH registers MCP servers through cordis.patch.yml, but the configuration is buried in YAML fields with no visual management surface and no quick way to tell which endpoints are actually reachable, which are stuck on authentication, and which have a protocol-version mismatch. Every triage session means re-reading the patch file and writing a throwaway probe script, which is slow and error-prone.

This static DSH plugin adds an MCP Services section under Settings: one card per configured server showing transport type, command or URL, auth headers, and a live status badge (connected / reachable / error / unknown). You can add, edit, or delete entries and every change persists straight back into cordis.patch.yml. The key feature is the probe button, which does not issue a bare TCP ping but performs a real MCP initialize handshake against each server: for streamable-http it fires an authenticated POST session through the official MCP SDK; for stdio it spawns the server process and completes the initialize round-trip over the pipe, trusting only the protocol version and server name the endpoint actually returns.

Well suited for developers running multiple MCP endpoints under DSH (file-system, database, browser tooling, etc.) who need to triage connectivity failures or protocol incompatibilities quickly, instead of re-reading YAML or writing throwaway probe scripts each time. Once installed, restart DSH and the new management page appears under Settings with zero manual configuration beyond the initial install.