{"version":"2.1.310","anchor":"mcp-claudeai-servers-toolpermissions-transform-and-standar","canonical_anchor":"mcp-claudeai-servers-toolpermissions-transform-and-standar","heading":"Some MCP servers handle tool permissions and large results differently","tier":"notice","area":"MCP","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.310\/e\/mcp-claudeai-servers-toolpermissions-transform-and-standar","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.310","markdown":"### Some MCP servers handle tool permissions and large results differently\n\nFor some MCP servers, Claude Code now reworks tool permissions, skips the overall tool-result size budget and tracks a standard key for grants\n\n**Unclear.** It is not clear which MCP servers this applies to.\n\n**What**\n\nMCP servers are programs that give Claude extra tools. For some of them, the way Claude Code lists their tools has changed:\n\n- The tool permissions the server sends are transformed before use, rather than used as they arrive.\n\n- These servers skip the overall size budget that limits how much tool output is kept.\n\n- Each tool gets a standard key for permission grants, along with a note of whether an older grant was refused.\n\nHosted MCP servers also get a new way of running tool calls whose results are not meant to be shown to Claude.\n\n**Why**\n\nThis looks like groundwork for how permission grants work with MCP tools. Large results from these servers are no longer cut back by the shared size budget.\n\n- Flag `tengu_mcp_strip_trailing_xml_tags`: Not enough to say (read for one account on one subscription tier against v2.1.310; this account: no value returned, anonymous baseline: no value returned, compiled default: off) These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release.\n- Area: MCP\n- Tier: You'll notice\n- Useful: 2\/5\n- Signal: 3\/5\n- Scope: individual\n- Heads-up: no"}