Hi everyone! I've spent the last few years building production AI agents, including an SRE agent that investigates incidents. The most painful part was always the same: a run fails, and you scroll through hundreds of lines of trace to find where it went wrong. Was it a loop? A broken tool? Did it just give up? Crescent Trace does that first pass for you in under a second. Paste a trace (LangGraph, OpenAI messages, OpenTelemetry/Langfuse, or JSONL), and it shows: - the exact step where your agent got stuck - what kind of failure it was (loop, retry storm, crashed tool, step limit) - how many tokens the loop wasted - what to fix It runs entirely in your browser, so your traces stay private. No signup, no API key. Try the "Retry storm" example to watch the dragon walk the trail and stop at the loop. It catches mechanical failures, not reasoning mistakes, so I'd love your feedback, especially traces it gets wrong. What should it detect next? Github: https://github.com/nischita44/cr...

Crescent Trace

Find where your AI agent got stuck, in your browser

Find where your AI agent got stuck, in your browser