{"version":"2.1.311","anchor":"mcp-serve-tool-meter-adds-started-mcp-servers-and-loop-stall","canonical_anchor":"mcp-serve-tool-meter-adds-started-mcp-servers-and-loop-stall","heading":"MCP server mode now reports how many servers started and event-loop delays","tier":"internal","area":"Telemetry","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.311\/e\/mcp-serve-tool-meter-adds-started-mcp-servers-and-loop-stall","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.311","markdown":"### MCP server mode now reports how many servers started and event-loop delays\n\nWhen Claude Code runs as an MCP server, its tool-call usage data now includes the number of started MCP servers and event-loop stall time\n\n**What**\n\nClaude Code can run as an MCP server, a program that offers its tools to other apps through the Model Context Protocol. In that mode, the usage information attached to each tool call now includes two more figures:\n\n- `started_mcp_servers`: how many MCP servers were started\n\n- `loop_stall_ms`: how long, in milliseconds, Claude Code's internal event loop was held up, measured by a regular timer that adds up any delays\n\n**Why**\n\nThis gives a clearer picture of how responsive Claude Code is while serving tools, and how many servers it is running.\n\n- Area: Telemetry\n- Tier: Under the hood\n- Useful: 1\/5\n- Signal: 1\/5\n- Scope: individual\n- Heads-up: no"}