Kōan - See your AI agents think. Reasoning, tool calls & decisions

Kōan

•13d ago

AI agents are powerful but invisible. When they fail, you get an error — not a reason. Kōan streams every agent event live — reasoning, tool calls, decisions, errors — tagged and timed. → Anthropic, OpenAI, Gemini + any custom LLM → Add DeepSeek, Mistral, Ollama via custom provider → Per-agent model selection → BYOK — keys never stored → No signup needed Free to start. Just add your API key and run.

Replies

Best

Kōan

Maker

📌

Hey Product Hunt! 👋 I built Kōan because I kept hitting the same wall — AI agents failing in production with zero visibility into why. The name comes from Zen Buddhism. A kōan is a riddle with no obvious answer — which is exactly what debugging a black-box AI agent feels like. Kōan makes the invisible visible. A few things I'd love your feedback on: 1. Which provider integrations matter most to you? 2. What would make the event stream more useful for debugging? 3. Would team sharing / log export unlock this for your workflow? Try it free at k-an.vercel.app — no signup, just add your API key and hit Run All. Happy to answer any questions below! 🎋

Report

14d ago

Kōan

Maker

Update — Day 3 🎋

Shipped a few improvements based on your feedback:

→ Clearer positioning: Kōan is agent observability — not just a log viewer or debugger.

→ JSON + CSV export: download full agent sessions for offline debugging, sharing, or analysis.

→ Session sharing: generate a URL and share exactly what your agent did — reasoning steps, tool calls, and decisions included.

→ Clearer empty state: new users now know exactly what to do when they land.

Kōan works with OpenAI, Anthropic, Gemini, DeepSeek, Mistral, Ollama, and custom endpoints.