{"version":"2.1.296","anchor":"large-mcp-results-skipped-in-subagents-new-shared-store","canonical_anchor":"large-mcp-results-skipped-in-subagents-new-shared-store","heading":"Large tool result cutting reworked: shared store, subagents skipped","tier":"notice","area":"MCP","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.296\/e\/large-mcp-results-skipped-in-subagents-new-shared-store","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.296","markdown":"### Large tool result cutting reworked: shared store, subagents skipped\n\nLarge MCP results are no longer cut inside subagents, cut text moves to a shared store, and the restore step no longer assumes MCP\n\n**What**\n\nWhen a tool from an MCP server (an outside tool provider connected to Claude Code) returns a very large result, Claude Code can cut it down and keep the full text so it can be fetched later. That mechanism changed:\n\n- Inside a subagent (a helper agent started by the main one), the result is now returned whole instead of being cut.\n\n- Elsewhere, the full text now goes into a shared store through a new helper, instead of a per-session store limited by count and size that could drop waiting entries.\n\n- The step that restores a cut result no longer hard-codes `mcp_large_result_cut` as its name. It reads the label carried by the cut itself, which suggests the mechanism may be used for more than MCP results.\n\n**Why**\n\nA subagent no longer receives a cut result it has no way to fetch in full.\n\n- Area: MCP\n- Tier: You'll notice\n- Useful: 2\/5\n- Signal: 1\/5\n- Scope: individual\n- Heads-up: no"}