The latest articles on DEV Community by Param Shah (@param_shah_e2b). https://dev.to/param_shah_e2b https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3879310%2F4f2140ed-1169-4194-ba87-e37fa6e87f21.png https://dev.to/param_shah_e2b en Param Shah Sat, 18 Apr 2026 11:27:11 +0000 https://dev.to/param_shah_e2b/we-built-traceai-an-open-source-tool-for-tracing-llm-calls-in-production-4b3m https://dev.to/param_shah_e2b/we-built-traceai-an-open-source-tool-for-tracing-llm-calls-in-production-4b3m <p>If you have ever tried to debug an LLM app in production, you know <br> how painful it gets. You have no idea what prompt actually went out, <br> what the model returned, how long it took, or why it failed.</p> <p>That is exactly why we built traceAI.</p> <p>traceAI is an open-source observability tool that traces every LLM <br> call in your application. It captures:</p> <ul> <li>Inputs and outputs</li> <li>Latency and token usage</li> <li>Costs</li> <li>Errors and failures</li> </ul> <p>All with minimal setup.</p> <p>We are launching our full platform next week but the traceAI repo <br> is already live on GitHub.</p> <p>Check it out: <a href="https://github.com/future-agi/traceAI" rel="noopener noreferrer">https://github.com/future-agi/traceAI</a></p> <p>Would love feedback from devs who are running LLMs in production. <br> What does your current observability stack look like? What is <br> missing?</p> ai llm opensource devtool