Login

The LLM gateway that learns to skip the LLM: trains its own tiny decision model on your agent's traffic and answers repeat decisions in 0.2 ms with no LLM call. OpenAI-compatible, one Rust binary. The server binary for pip users: pankhllm serve