Trackly - Zero-overhead AI cost and usage tracking

While building agentic AI systems with LangChain, I kept hitting the same wall: I had no idea what each feature was actually costing me. Was the RAG pipeline burning my budget? Was Groq cheaper than OpenAI for my workload? I couldn't tell. And existing solutions either required a proxy, added latency, or were too complex to set up. So I built Trackly - a zero-overhead Python SDK that wraps your existing LangChain models and automatically tracks tokens, cost, latency, and metadata with just 2 lines of code. No proxies. No refactoring. No surprises on your bill.