Currently when an MCP server connection fails or terminates, there's no built-in way to reconnect without restarting the Claude CLI session. This creates friction in the development workflow when a
server connection drops.

Steps to reproduce:

1. Start a Claude Code session with MCP servers
2. Observe one or more servers showing "failed" status via mcp command
3. Attempt to reconnect the failed server

Current behavior:

• No command exists to reconnect a failed MCP server
• User must restart the entire Claude CLI session

Expected behavior:

• Add a command like mcp reconnect [server-name] to restart specific failed servers
• Alternatively, add a --reconnect flag to the existing mcp command

Impact:
Medium - This is a workflow improvement that reduces interruptions when server connections fail temporarily.

Additional notes:

• The playwright server appears to fail more frequently than others
• A UI indicator when a server disconnects would also be helpful