Tighter time budgets and response caps so tool calls land inside Claude Desktop's window — and when a tool does run out of time, Claude now offers you a Continue instead of breaking the conversation.
What shipped
•Tool time budgets tightened across the board. Most tools now finish in 45s or return a clean timeout, well inside Claude Desktop's tool-call window — the "Tool result could not be submitted" error banner should appear far less often.
•Response size cap halved from 200KB to 100KB. Large Braze pulls (instance audits, full Canvases, template lists) now trim gracefully to fit rather than silently overflowing Claude's tool-result submission limit.
•When a tool times out, Claude now offers a Continue prompt. Instead of dead-ending the conversation with an "expired" banner, the tool returns a structured response that tells Claude to ask: "That request hit the time budget. Want me to retry with a narrower scope?" Answer yes and Claude reruns with reduced inputs.
•Truncated responses now include a "continue hint" telling Claude to surface the trimming + offer to fetch the rest with a narrower query — so you never silently lose data because it didn't fit.