{"version":"2.1.295","anchor":"mcp-failed-server-redial-reworked","canonical_anchor":"mcp-failed-server-redial-reworked","heading":"Reconnecting to failed MCP servers reworked","tier":"notice","area":"MCP","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.295\/e\/mcp-failed-server-redial-reworked","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.295","markdown":"### Reconnecting to failed MCP servers reworked\n\nRetries for MCP servers that failed to connect now report a result for each try, and one failed server can be retried on its own\n\n**What**\n\nMCP (Model Context Protocol) servers are outside programs that give Claude extra tools. When one fails to connect, Claude Code tries again. That retry has been reworked:\n\n- each attempt now reports an outcome instead of running in a single loop with no result\n\n- a specific failed server can be retried on its own, if that is still allowed\n\nRequests from servers asking you for input are now also skipped for some servers.\n\n**Why**\n\nIn practice, reconnecting is handled one server at a time with a clear result, which should make it easier to recover a single server that dropped.\n\n- Area: MCP\n- Tier: You'll notice\n- Useful: 3\/5\n- Signal: 2\/5\n- Scope: individual\n- Heads-up: no"}