Describe an intrusive urge and watch a Strands agent ground it against your own anchors, name the thinking trap, and guide an Exposure & Response Prevention delay — with its whole chain-of-thought drawn live.
Watch the agent reason in real time — each tool call appears as a node, streamed over SSE.
Objective facts you write while calm. Every urge is grounded against them, not guessed at.
For compulsive urges the agent pauses and guides a timed delay — you sit with it before acting.
A model-driven Strands agent on Amazon Bedrock (Claude), with a credential-free demo mode.
Every lifecycle event is streamed back over SSE and drawn on the map as it happens.
No sign-up. Runs on a built-in demo model out of the box, or plug in your Amazon Bedrock key for a live LLM.