Every reasoning step, tool call, and decision — visualized as a live interactive graph. See what's really happening inside the black box.
The Problem
They read files, search the web, execute code, make decisions — but you only see the final answer. SYNAPSE cracks open the box.
Features
Run npm run bridge:claude-code and SYNAPSE tails the active session — every thought, tool call, sub-agent spawn, and file edit unfolds as a live graph. Or pipe any agent in over WebSocket.
4 pre-loaded sessions: multi-agent orchestration, website building, debugging, security audits. Full playback controls at 0.25–8× speed.
Drop a Claude Code .jsonl session or any OpenClaw transcript. Real conversation DAG via parentUuid — sub-agents land in their own swim lanes.
Visualize orchestrators spawning sub-agents. Color-coded paths show which agent handles what — from research to writing to review.
The Story
Data (the AI) designed the architecture, chose the tech stack, wrote every line of code, and made all design decisions — autonomously. The human provided the vision and the deadline. The AI did the rest.
~12 hours of actual AI development time across 6 working days. Built with Next.js, React Flow, Tailwind CSS, Framer Motion, Pusher, and OpenClaw. Deployed on Vercel. Search powered by self-hosted SearXNG.
Explore demo sessions or connect your own agent. This is what AI reasoning looks like.
Try SYNAPSE