Upsolve AI
Day 25 after signup
π³ See every step your data agent took
Traces, SQL and live evals for every conversation, so you catch wrong answers early.
Why it works
- Lead with a clear summary of the capability so readers can quickly understand what is new.
- Explain the problem before showing how the feature addresses it, then support the explanation with concrete capabilities.
- Give existing and prospective users distinct next steps so each audience has a relevant action.
How it is built
- Newsletter introduction
- Feature summary
- Problem statement
- How it works
Email copy
Hey there, Welcome back to your Upsolve AI newsletter! More people asking your agent questions is good news, but only if you can see what it tells them. This issue covers the view that shows you. TL;DR Every conversation, trace, SQL query and skill call your agent makes, in one place. Live LLM-as-judge evals flag wrong answers, lazy responses, empty results and frustrated users. Embedded, Slack and Claude conversations all land in the same history. π See every step your data agent took, and fix it. A slightly wrong number looks exactly like a right one. Users who get a bad answer ask again, then quietly stop using the agent. Without traces, builders can't tell whether the problem was the data, the context or the question. π οΈ How it works See every session. The agent chat history shows every user, question, eval result and cost. Trace every step. Open any session to see RAG retrieval, the skill used, the SQL, validation and the chart. Close the loop. When an eval flags a problem, fix the prompt, skill or data model and watch the next conversations improve. βΆ Watch Serguei's full demo π¦ What's included Live evals on real usage: answer correctness, agent laziness, null or empty results, chart failures, user frustration or create your own evaluation criteria. Full step-by-step traces: the exact SQL, RAG sources, knowledge base, connected MCP and tool calls behind every answer. Cost per trace: see spend, charts generated and errors for each individual trace. One view across surfaces: embedded, Slack and Claude conversations all land in the same history. π Try it Already a customer? Observability is live in your workspace. Open your chat history, pick one answer an eval flagged, and trace it back to its cause. New to Upsolve? Start free at upsolve.ai β with 2,000 AI credits, or read the full feature page. Next issue: the Data Plane, which brings Stripe, PostHog and HubSpot into one agent with real SQL. Questions? Reply to this email or reach us at kaling@upsolve.ai Follow the series on YouTube β Subscribe to Upsolve AI Building the future of data-intelligent AI, The Upsolve Team The conduit between data and decisions Stay tuned for more exciting updates, customer stories, and news from us! - Team Upsolve AI You are receiving this email because you opted in to receive email from the list Product Updates & Releases from Upsolve AI Upsolve AI, 1111B S Governors Ave STE 6587 Dover, DE 19904 Unsubscribe