Telemetry-aware AI agents that act in Slack, Linear, and GitHub, without your production data leaving your cloud

Most AI observability assumes you'll ship your telemetry to a vendor's cloud so their model can reason over it. If your data is sensitive or high-volume, that's a non-starter. So we did it the other way around: the agent runs where the data already lives. We're building groundcover, and we're spending an evening in SF showing how the new expansion to Agent Mode actually works. The short version: agents that read live production telemetry and act across your dev toolchain (Slack, Linear, GitHub) through remote MCP connectors, every action attributable to a specific user. The main event is a live demo. We trigger a real anomaly on stage, Agent Mode ingests the telemetry, isolates the cause, and spins up an agentic workflow that opens a patch through GitHub. No recorded run, no slideware. We'll also walk through how custom Skills map LLM-driven logic to a real infrastructure runbook, and how MCP authorization keeps every agent action tied to a user and fully revocable, all inside your own cloud under BYOC. Food, drinks, and time to talk. The deep architecture questions, we'll connect…