{"version":"2.1.292","anchor":"workflow-agents-per-agent-model-tracking-and-rate-limit-mod","canonical_anchor":"workflow-agents-per-agent-model-tracking-and-rate-limit-mod","heading":"Workflow agents track model changes during a run","tier":"internal","area":"Workflows","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.292\/e\/workflow-agents-per-agent-model-tracking-and-rate-limit-mod","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.292","markdown":"### Workflow agents track model changes during a run\n\nWorkflow agents now pass an admission step before starting, follow model changes mid-run, and report the model on rate-limit errors\n\n**Unclear.** What the new admission step checks before a workflow agent starts is not known.\n\n**What**\n\nAgents started by a workflow (a multi-step task Claude Code runs on your behalf) have changed in three ways:\n\n- they go through an admission step before they start\n\n- they follow changes to the model they are using during a run, instead of keeping the model chosen when they were started\n\n- rate-limit errors from them now say which model was in use\n\n**Why**\n\nWhen a workflow agent switches model partway through, for example to a fallback model, usage and rate-limit errors are tied to the model that was actually used.\n\n- Area: Workflows\n- Tier: Under the hood\n- Useful: 1\/5\n- Signal: 2\/5\n- Scope: individual\n- Heads-up: no"}