Group of 3 Use it now Notable
A new gated 'advisor' tool was added, restricted by an env var and account entitlement, with clearer in-context messaging about when it applies
What
A new internal tool/event kind called "advisor" has been introduced.
- It can be disabled with an environment variable, and otherwise requires a first-party Anthropic account plus an internal entitlement check before it's available.
- The availability check was refactored into a shared helper function, replacing inline checks that combined the disable env var, first-party account check, and entitlement check.
- When you change the advisor model setting mid-conversation (from the
/advisorUI or its slash command), Claude Code now shows a contextual note explaining when the new setting will actually take effect — for example, if the API already refused to use the advisor this turn, or if the conversation already locked in a different advisor model or tool declaration.
Why
This lays the groundwork for an advisor feature restricted to eligible accounts, and helps avoid confusion by telling you explicitly when a mid-conversation change to the advisor setting won't apply immediately.
Names in the bundleCLAUDE_CODE_DISABLE_ADVISOR_TOOL/advisor
Documented inclaude-code/advisor
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
Confirmed since
Anthropic's documentation has since written up CLAUDE_CODE_DISABLE_ADVISOR_TOOL, on Connect Claude Code to an LLM gateway.
| `400` errors on every request, in the gateway's own words rejecting an unrecognized tool type, such as `Input tag 'advisor_20260301'`, on Claude Code v2.1.275 | In a gradual rollout on that version, requests carry an [advisor tool](/docs…llm-gateway-connect see the edit
Confirmed since
Anthropic's documentation has since written up /advisor, on Connect Claude Code to an LLM gateway.
| `400` errors on every request, in the gateway's own words rejecting an unrecognized tool type, such as `Input tag 'advisor_20260301'`, on Claude Code v2.1.275 | In a gradual rollout on that version, requests carry an [advisor tool](/docs…llm-gateway-connect see the edit
Added since
A small documentation edit on optimizing-for-cost-and-intelligence touched a line naming /advisor after this was published.
In agent loops the cache-read term is usually the largest of the five; if not, check that caching is engaged. When the [advisor tool](https://platform.claude.com/docs/en/agents-and-tools/tool-use/advisor-tool#usage-and-billing) or [compact…about-claude/models/optimizing-for-cost-and-intelligence see the edit
One source agreesOne thing we can check says the same as this entry.
Anthropic's documentation agrees
A small documentation edit on optimizing-for-cost-and-intelligence touched a line naming /advisor after this was published.