fix: keep disabled state across in-flight reconnect results

A reconnect request that completes after the server was disabled no
longer overwrites the disabled state. Connection attempts fold into
disabled when the config no longer allows the server, closing any
fresh connection and clearing its caches; the mcp_toggle enable and
mcp_reconnect handlers apply the same check before writing their
results, and the reconnect failure path keeps disabled when the server
was disabled mid-flight.

src/services/mcp/useManageMCPConnections.ts

@@ -325,6 +325,21 @@ export function useManageMCPConnections(
       commands: Command[]
       resources?: ServerResource[]
     }) => {
+      // A connect that completes after the server was disabled (e.g. disable
+      // during a reconnect request) must not resurrect the server: close any
+      // fresh connection, clear its caches, and mark it disabled instead.
+      if (isMcpServerDisabled(client.name)) {
+        if (client.type === 'connected') {
+          void clearServerCache(client.name, client.config)
+        }
+        updateServer({
+          name: client.name,
+          type: 'disabled',
+          config: client.config,
+        })
+        return
+      }
+
       updateServer({ ...client, tools, commands, resources })
 
       // Handle side effects based on client state
@@ -1027,7 +1042,11 @@ export function useManageMCPConnections(
               `Max reconnection attempts (${MAX_RECONNECT_ATTEMPTS}) reached, giving up`,
             )
             reconnectTimersRef.current.delete(name)
-            updateServer({ ...client, type: 'failed' })
+            if (isMcpServerDisabled(name)) {
+              updateServer({ name, type: 'disabled', config: client.config })
+            } else {
+              updateServer({ ...client, type: 'failed' })
+            }
             return
           }
         }