Show HN: Makoto – blocks AI agent claims that contradict its logged actions

- Makoto is an open-source tool that audits AI agents by comparing their natural-language claims ("I sent the email") against a structured log of their actual tool calls and API actions. - It works by taking a JSON action log and a user-provided claim, then using an LLM to reason step-by-step whether the claim matches the logged events — flagging contradictions or hallucinations. - Many AI agents now generate post-hoc summaries of what they did; these summaries can be inaccurate or fabricated. Makoto addresses the growing need for verifiability and guardrails in agentic workflows. - The project is by Clear-Sights and was posted on Hacker News as a "Show HN" — a category for sharing something you built.