{"version":"2.1.307","anchor":"agent-summary-can-use-a-configured-writer-model-fableopus","canonical_anchor":"agent-summary-can-use-a-configured-writer-model-fableopus","heading":"Background agent summaries can be written by a separate model","tier":"notice","area":"Subagents","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.307\/e\/agent-summary-can-use-a-configured-writer-model-fableopus","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.307","markdown":"### Background agent summaries can be written by a separate model\n\nShort summaries of background agents can now come from a different writer model, set by the server, when the agent runs on fable or opus\n\n**Unclear.** Which writer model, if any, the server names is not known.\n\n**What**\n\nClaude Code writes short summaries of what a background agent is doing. A background agent is a subagent, a helper Claude starts, working on a task out of view. These summaries can now be written by a separate writer model instead of the agent's own model. This only happens when all of the following hold:\n\n- the server's summary settings name a writer model\n\n- the summary is the short kind\n\n- the agent's own model belongs to the fable or opus family\n\nIf a call to the writer model fails, Claude Code stops using it for the rest of the run. Usage reports about skipped and stopped summaries now also record the writer model, the configured writer, and whether the writer failed.\n\n**Why**\n\nA different model writing the summaries can change what they cost and how good they are. When the server names no writer, the agent's own model is used as before.\n\n- Area: Subagents\n- Tier: You'll notice\n- Useful: 4\/5\n- Signal: 4\/5\n- Scope: individual\n- Heads-up: no"}