{"version":"2.1.321","anchor":"mcp-reconnect-decision-considers-config-readability-and-poli","canonical_anchor":"mcp-reconnect-decision-considers-config-readability-and-poli","heading":"Claude Code stops retrying MCP servers blocked by policy or unreadable setup","tier":"notice","area":"MCP","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.321\/e\/mcp-reconnect-decision-considers-config-readability-and-poli","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.321","markdown":"### Claude Code stops retrying MCP servers blocked by policy or unreadable setup\n\nFailed MCP servers are no longer retried when their setup cannot be read or a policy blocked them\n\n**What**\n\nMCP servers are outside programs that give Claude extra tools. Claude Code regularly decides whether to reload your MCP server setup, retry servers that failed, or do nothing. Before, any failed server led to a retry. Now:\n\n- A failed server is retried only if its setup could be read and it was not blocked by policy.\n\n- A connected server whose setup was refused makes Claude Code reload the setup.\n\n**Why**\n\nClaude Code no longer keeps retrying servers that cannot connect because a policy blocks them, which would fail every time.\n\n- Area: MCP\n- Tier: You'll notice\n- Useful: 2\/5\n- Signal: 1\/5\n- Scope: individual\n- Heads-up: no"}