{"version":"2.1.246","anchor":"killing-an-mcp-task-now-tears-down-its-helper-session-too","canonical_anchor":"killing-an-mcp-task-now-tears-down-its-helper-session-too","heading":"Killing an MCP task now tears down its helper session too","tier":"notice","area":"MCP","scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.246\/e\/killing-an-mcp-task-now-tears-down-its-helper-session-too","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.246","markdown":"### Killing an MCP task now tears down its helper session too\n\nKilling an MCP task now cleans up its helper session and metadata instead of leaving state behind.\n\n**What**\n\nKilling an MCP task aborts the drive controller and issues a cancel alongside the main abort, clears the parked state, and waits for the pending write to finish before deleting the helper session and project directory metadata. Less chance of leftover state after a kill.\n\n**Details**\n\n- Runs unconditionally for MCP task entries.\n\n**Evidence**\n\n`McpTask.kill deleteMcpTaskMetadata: `\n\n- Area: MCP\n- Tier: You'll notice\n- Useful: 3\/5\n- Signal: 2\/5"}