Meta-Reasoning is an SDK for governing how LLMs reason. Instead of treating models as agents, it externalizes cognition into a controllable architecture: observe reasoning structure, mutate it with constraints, and record trajectories in an epistemic ledger. Enable debugging, replay, benchmarking, and CI/CD for reasoning. Make LLM behavior testable, deterministic, and governed.
Already available on 🦞OpenClaw, 🟠ClaudeCode and 🤖Codex!
