The advisor is a tool that lets the model doing the work consult a stronger model when it is stuck or facing an important decision. A fast, cheaper model handles the routine steps, and a higher-intelligence model steps in only at the moments that matter — aiming for near top-model quality at a lower total cost. This guide covers the /advisor command in Claude Code and the API-side concept, based on the official docs.
🟢 Current lineup (September 2026): Claude Opus 5.5 / Claude Sonnet 5 / Claude Haiku 4.5 (higher tier: Claude Fable 5.1). · model notice · Fable subscription
Fable 5 and 5.1 subscription (updated September 7, 2026): Claude Fable 5.1, released September 1, 2026, is the current Fable model and Fable 5 is now legacy. Plan terms are the same for both — Max and Team Premium plans include Fable at up to 50% of the weekly usage limit; Pro and Team Standard use usage credits (
What the advisor tool is
Per the official docs, the advisor tool lets the executor — the model actually reading files and editing code — call a stronger advisor model at key moments: before committing to an approach, when an error keeps recurring, and before declaring a task complete. The advisor receives the full conversation, including every tool call and result, and returns guidance that the executor applies before continuing. It runs server-side on Anthropic's infrastructure as a server tool.
Requirements
- It is an experimental feature; the official docs state that behavior, pricing, and availability may change.
- It requires Claude Code v2.1.98 or later (run
claude update). - It works only via the Anthropic API — not on Amazon Bedrock, Google Cloud's Agent Platform, or Microsoft Foundry.
Turning it on and off — /advisor
Run /advisor in Claude Code and pick the model to use as the advisor — a typical pairing is Sonnet executing with Opus advising. To stop, run /advisor off or choose "No advisor" in the picker. To disable it entirely (for example, org-wide), set the environment variable CLAUDE_CODE_DISABLE_ADVISOR_TOOL=1; the /advisor command then becomes unavailable and any configured advisorModel is ignored.
Supported model pairings
Per the official docs, supported main models are Opus 4.6 or later, Sonnet 4.6 or later, or Haiku 4.5, and Fable 5 also qualifies on Claude Code v2.1.170 or later. Its successor Fable 5.1 requires Claude Code v2.1.255 or later; check the official docs for whether it is supported in advisor pairings. If either the main model or the advisor is a model Claude Code does not recognize, the advisor is not attached. Any accepted pairing works; different pairings trade cost against capability differently.
When does it consult? The model decides
Advisor timing is model-driven, not rule-based. The docs say it tends to consult before committing to an approach, when an error keeps recurring, and before declaring a task done. If you want more or fewer consultations, say so in your prompt — you can also request one directly, for example "consult the advisor before you continue." There is no setting to cap or force advisor calls. While a consultation is in progress, the transcript shows an "Advising" line with the advisor model's name.
Cost and usage
Because the advisor is called at decision points rather than on every turn, the official docs note that pairing a faster main model with a stronger advisor typically costs less than running the stronger model throughout. Advisor usage counts toward the session totals shown by /usage. Also worth knowing: toggling /advisor — unlike changing model or effort level — does not invalidate your main model's prompt cache.
Using it from the API
The same capability is available in beta on the Claude API (Messages API): include the beta header advisor-tool-2026-03-01 and declare the advisor in your tools array. Full details are in the official API docs; for the Claude Code side, see the official advisor page.
For broader Claude Code cost management, see our Claude Code pricing guide, and for tool-use fundamentals, the Tool Use guide.
Sources: the official Claude Code docs (code.claude.com/docs/en/advisor) and Claude API docs (platform.claude.com). As an experimental/beta feature, behavior, pricing, and availability may change.
Related articles
- Claude Code Use Cases: Explore, Debug, Refactor, Test, and PR Workflows
- Claude Code GitHub Integration — Automate PRs and Reviews with @claude
- Claude Code Slash Commands — Built-in Commands and Custom Ones
- Claude Code Subagents — Specialized AI to Isolate Side Work
- Claude Code Hooks — Rules That Run Automatically Every Time