Light in Claw sits between an agent and its output, illuminating every decision along the way.
Any goal in plain language. The task is handed to an OpenClaw-based agent that decomposes it into a plan of actions and sub-decisions.
As the agent works, Hermes Nous narrates each step: what it's considering, why it chose one path over another, what it's uncertain about. This streams live to the console, not as an afterthought.
The complete ordered reasoning trace is serialized and SHA-256 hashed. The hash is a fingerprint — change one word of reasoning and the hash changes. It can be anchored on Solana for a permanent, tamper-evident record.
Each illumination gets a shareable permalink. Re-hash the trace yourself and compare. The proof travels with the conclusion — trust becomes verifiable instead of assumed.
Agent runtime & skill framework.
Reasoning narration via OpenRouter.
Low-latency backend on Railway.
Trace storage & permalinks.
On-chain proof anchoring.