{"version":"2.1.287","anchor":"mcp-remote-server-retry-reworked-with-telemetry","canonical_anchor":"mcp-remote-server-retry-reworked-with-telemetry","heading":"Remote MCP servers that fail at startup are retried and picked up more reliably","tier":"notice","area":"MCP","scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.287\/e\/mcp-remote-server-retry-reworked-with-telemetry","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.287","markdown":"### Remote MCP servers that fail at startup are retried and picked up more reliably\n\nClaude Code now records each retry of a remote MCP server that failed to connect, and uses servers that recover as soon as they connect\n\n**Unclear.** It is not clear what the wait between retries is now tied to.\n\n**What**\n\nMCP servers (Model Context Protocol servers, which give Claude Code extra tools from outside programs or services) can be remote, reached over the network. When a remote MCP server fails to connect, Claude Code tries again. Each of those retry attempts is now recorded on its own. A server that recovers on a later attempt is now picked up as soon as it connects.\n\n**Why**\n\nIf a remote MCP server is briefly unavailable when you start Claude Code, its tools are more likely to become available once it comes back.\n\n- Area: MCP\n- Tier: You'll notice\n- Useful: 3\/5\n- Signal: 1\/5"}