docs: advisor tool — workhorse mode (sonnet executor + opus/fable advisor)#126
Merged
Conversation
…he settings schema Claude Code natively supports the API advisor tool (beta) — a cheaper executor session consults a stronger advisor model mid-turn, server-side, with the full transcript. It works on Max subscription billing, which makes it a third routing lever alongside the Opus/Sonnet/Codex split. Recommended as opt-in "workhorse mode" (/model sonnet + /advisor opus, or fable while access lasts) rather than a composed default: our standing model is opus[1m] where an advisor is marginal, and a Fable session only accepts a Fable advisor.
4 tasks
arzafran
added a commit
that referenced
this pull request
Jul 8, 2026
Both upstream versions (2.1.203/2.1.204) are bug-fixes-only — background-agent reliability, TUI polish, headless SessionStart hook streaming — so this release is manifest/docs tracking only, no schema or config changes. Adopted: - Track CLAUDE_CODE_DISABLE_MOUSE (full mouse-capture opt-out, companion to _CLICKS) in manifest knownEnvVars + docs env table - Manifest housekeeping: advisorModel added to knownSettingsKeys, closing the false-positive scanner warning left by PR #126 Deletions: - none (checked the 2.1.203 permission-mode footer badge against statusline.ts — no overlap)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this does
Documents Claude Code's native advisor tool for the team and makes
advisorModela known settings key. The advisor lets a cheap Sonnet session consult Opus or Fable mid-turn with the full transcript — on Max subscription billing — so daily-driver sessions get near-top-tier planning while burning mostly Sonnet quota, saving scarce Opus hours for work that needs a top-tier executor.Summary
docs/agent-models.md: new "Advisor" section — config surface (/advisor,advisorModel,--advisor,CLAUDE_CODE_DISABLE_ADVISOR_TOOL), pairing table (sonnet executor is the sweet spot; a Fable session only accepts a Fable advisor), recommended opt-in "workhorse mode", and interactions with subagent inheritance, Codex routing, and Fable's encrypted adviceMANUAL.md: user-facing note in Advanced, next to the model/platform sections (advisor is Anthropic-API-only, like the Bedrock/Vertex caveats above it)src/schemas/settings.ts+ regeneratedschemas/settings.schema.json: addadvisorModelso a persisted/advisorchoice validates like every other known keyDeliberately not a composed default: standing model is
opus[1m], where an advisor is marginal.Test Plan
bun run typecheckcleanbun test— 777 passbun run lint— no new findings (4 pre-existing warnings untouched)bun run schemas:emitregenerated and committed