Anthropic's official advisor tool: Opus on call, Haiku swarms in parallel, Sonnet drives Claude Code
aitrendz_xyz · x · 2026-10-08
Anthropic's docs detail the experimental advisor tool for Claude Code: the main model consults a stronger model at key moments (before plan lock-in, on recurring errors, before completion), with full conversation visibility, available to subscription and API accounts.
The recommended setup: claude --advisor opus --subagents haiku — Sonnet 5.5 drives the session writing diffs and running tests, Haiku 5.5 subagents swarm the repo at 340 tok/s for file discovery and docs lookup, and Opus 5.5 steps in only at three critical checkpoints. Best for large refactors and stubborn debugging.
More from coding & agent
- Enterprise agent lesson: permissions don't guarantee correct business transactions — shashib · 2026-10-08
- Fintech Rebuilt Its Data Foundation in 15 Months—AI Never Touched Production — alex_verem · 2026-10-08
- Meta Data: AI-Reviewed Diffs Have 1/50th the Production-Incident Rate of Human Review — arjunrajlab · 2026-10-08
- Opus 5.5 Lands on Grok Bot—How to Build a 24/7 AI Quant Research Desk — dean_rie · 2026-10-08
- RSIGym treats recursive self-improvement as systems engineering; Opus 5 leads at 0.4809 — teortaxesTex · 2026-10-08
- Codex Hits 40M Weekly Users, VraserX Predicts It Will Overtake Claude Code Soon — VraserX · 2026-10-08