High performance boost
Boost AI efficiency & token savings
Eliminate conversational bloat, cut latency, and reduce your monthly OpenAI and Anthropic bills with a single line of code.
Turbo AST compression
Compresses input prompts and RAG contexts before reaching target models, stripping redundant delimiters and natural language padding.
Up to 68%input reduction
Output brevity governor
Directs downstream models to omit conversational filler and verbose markdown wrappers, targeting expensive 3x to 5x output pricing.
Up to 50%output savings
Ultra-low edge latency
In-memory deterministic compression runs at the edge in sub-15ms, delivering faster time-to-first-token (TTFT) for user queries.
< 15msproxy overhead
Safe Vault syntax guarantee
Code fences, JSON schemas, variables, and markdown tables are shielded bit-for-bit without AST corruption.
100%syntax validity
Boost your existing API clients today
Simply update your baseURL to proxy.curtly.dev/v1 in standard SDKs.