{"version":"2.1.271","anchor":"max-output-tokens-recovery-now-surfaces-as-an-agent-mode-s","canonical_anchor":"max-output-tokens-recovery-now-surfaces-as-an-agent-mode-s","heading":"New 'recovering' status shown when a turn retries after hitting the output token limit","tier":"notice","area":"Response Handling","scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.271\/e\/max-output-tokens-recovery-now-surfaces-as-an-agent-mode-s","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.271","markdown":"### New 'recovering' status shown when a turn retries after hitting the output token limit\n\nThe spinner now shows a 'picking the thought back up' status while a turn retries after hitting the output token limit\n\n**What**\n\n- When a turn hits the output token limit and is being retried (reason `max_output_tokens_recovery`), the status line\/spinner now shows a new \"recovering\" state instead of the usual thinking\/tool text.\n\n- It displays as \"picking the thought back up\", or \"picking the thought back up (N of 3)\" on later retry attempts, with a limit of 3 attempts.\n\n- This is driven by a new `outputLimitRecovery` payload (attempt number and the limit) that's threaded through an agent-mode \"requesting\" event and into the spinner component's props.\n\n**Why**\n\nThis gives a clear, specific status message when Claude is automatically retrying after running out of output tokens, instead of leaving the status line looking stuck or generic.\n\n- Area: Response Handling\n- Tier: You'll notice\n- Useful: 3\/5\n- Signal: 2\/5"}