Building AI Agents That Actually Work at Work | Lessons from Tiger Data's Eon Agent

Tiger Data engineers John Pruitt and Justin Murray join developer advocate Jacky Liang to share how they built Tiger Agents at Work and Eon, an AI coworker that lives in Slack. They break down their architecture choices, why they built in-house, and how MCP, Postgres, and TimescaleDB power their internal agents. From observability to system prompts, this talk dives deep into what it takes to run production-grade workplace AI. 🛠 𝗥𝗲𝗹𝗲𝘃𝗮𝗻𝘁 𝗥𝗲𝘀𝗼𝘂𝗿𝗰𝗲𝘀 📌 Sign up for FREE ⇒ https://tsdb.co/td-yt 📌 Read the blog post ⇒ https://www.tigerdata.com/blog/we-built-production-agent-open-sourced-everything-we-learned 📌 Github link ⇒ https://github.com/timescale/tiger-eon 💻 𝗙𝗶𝗻𝗱 𝗨𝘀 𝗢𝗻𝗹𝗶𝗻𝗲! 🔍 Website ⇒ https://tsdb.co/homepage 🔍 Slack ⇒ https://slack.timescale.com 🔍 GitHub ⇒ https://github.com/timescale 🔍 Twitter ⇒ https://twitter.com/timescaledb 🔍 LinkedIn ⇒ https://www.linkedin.com/company/tigerdata 🔍 Tiger Data Blog ⇒ https://tsdb.co/blog 🔍 Tiger Data Documentation ⇒ https://tsdb.co/docs 📚 𝗖𝗵𝗮𝗽𝘁𝗲𝗿𝘀 ⏱ 0:00 ⇒ Introduction ⏱ 1:34 ⇒ Why build an AI Agent for Slack? ⏱ 6:28 ⇒ Dealing with API Limitations ⏱ 8:50 ⇒ The Role of Postgres and Time-Series ⏱ 12:29 ⇒ MCP and System Architecture ⏱ 15:17 ⇒ Interactions and Limitations with Eon ⏱ 21:00 ⇒ Using MCPs as 'Microservices' ⏱ 23:42 ⇒ Observability, Edge Cases, and LLM Errors ⏱ 28:24 ⇒ Lessons Learned and Key Takeaways ⏱ 34:50 ⇒ Outro