Last released May 22, 2026
Dynamic, on-the-fly generated harnesses for AI agents — the LLM writes the entire runtime (system prompt, loop, tools, schema, verifier) per task.
Supported by
