What
Requests that Claude Code sends to Anthropic's API (the service that runs the model) may now carry a header (an extra labelled line sent with a request) that lists the tools the session uses.
- The header is named
x-claude-code-sandy-ploverand is only attached for first-party addresses, meaning Anthropic's own API. - The list of tool names is capped at 7168 bytes. Common tools are placed last, and
incomplete_listis reported when names had to be left out. - Whether the header is sent depends on a feature flag,
tengu_snug_mitten, plus one other check. Nothing has been read about the state of that flag.
The same release also changes how requests are sent:
- Request bodies that are not plain text are now gzipped (compressed) too, using a helper that can handle streamed data. Before, only plain-text bodies were compressed. Extra size details about each body are now recorded.
- When estimating the token count of tool calls (a token is a small chunk of text the model reads), Claude Code now remembers the length of each call's input instead of measuring it again.
Why
This changes what is sent with each API request. When the header is in use, Anthropic can see which tools a session uses.
tengu_snug_mitten Not enough to sayNothing here resolved what this flag was doing on this version, so nothing here should be read as on or off.
This account: no value returned · anonymous baseline: no value returned · compiled default in v2.1.311: off
These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release.
Read once, for one account on one subscription tier, against v2.1.311. It isn't a statement about your account. What a flag value here can and cannot tell you
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
New in this build: tengu_snug_mitten