{"version":"2.1.314","anchor":"mcp-tool-wire-name-assignment-keeps-already-sent-names-stabl","canonical_anchor":"mcp-tool-wire-name-assignment-keeps-already-sent-names-stabl","heading":"MCP tool names stay the same once the model has seen them","tier":"notice","area":"MCP","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.314\/e\/mcp-tool-wire-name-assignment-keeps-already-sent-names-stabl","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.314","markdown":"### MCP tool names stay the same once the model has seen them\n\nClaude Code keeps the names of MCP tools it has already sent to the model instead of renaming them, and sorts out clashing names between servers\n\n**What**\n\nMCP servers are outside programs that give Claude Code extra tools. Claude Code gives each of these tools a name before it shows the tool to the model. It now keeps any name it has already sent to the model, so that tool is not renamed later in the conversation.\n\nIt also handles the case where two servers have tools whose names would end up spelled the same way.\n\n**Why**\n\nThe tool names the model sees stay the same for the whole conversation. Stable names also help the prompt cache, which is how Claude Code reuses work from earlier requests instead of sending everything again.\n\n- Area: MCP\n- Tier: You'll notice\n- Useful: 2\/5\n- Signal: 1\/5\n- Scope: individual\n- Heads-up: no"}