tool-interception

Models trust their tools blindly. OpenIntercept puts that trust boundary in your hands; an OpenCode plugin to hold, inspect, edit, spoof, or drop any tool call before the model sees it. Local-only, no telemetry.

Self-correcting intelligence layer for Agents using less capable models with OpenClaw - gives any model the resilience of frontier models