Anthropic’s Claude team announced on X that prompt cache diagnostics are now available in Claude Console, giving users a way to see “exactly which part” of a prompt changed when a request misses the cache, along with how many tokens that miss cost.
The update appears aimed at making cache misses less opaque for developers watching token usage. In the replies, several users described the feature as useful for debugging and cost tracking, with one commenter calling it a way to see “exactly where the tokens got cooked,” while another wrote that cache misses had often been “pure guesswork” before such diagnostics.
A few replies also raised practical questions about the rollout, including whether the feature applies to API usage only and whether subscription users will see the same transparency in the console. Others suggested the new view could help spot small prompt edits that quietly break cache hits, especially in longer prompts where the changed section may be easy to miss.
Source: ClaudeDevs on X
