BREAKING
Autocache auto-injects Claude cache headers
0
%
cost cut
0
%
latency cut
15,000-token agent: $0.045 to $0.0045
First call
0.045
Cached
0.0045
Drop-in proxy, no code changes
1
Run proxy on localhost:8080
↓
2
Repoint API base URL
↓
3
Headers injected auto
Three cache strategies
conservative
2 max
●
System/tools focus
moderate
3 max
●
Default, standard setup
aggressive
4 max
●
Maximum caching
Caching gains, if breakpoints fit
AI NEWS BLITZ
A new open-source proxy called Autocache brings Prompt Caching to Claude API calls automatically.