Schedule Tue 24 Ai + Ml
Ai + Ml Reviewed

Driving Adoption and Automation With MCP in Production at Liftoff

Session · Tuesday — Main Conference Day 1

13:30–14:00 UTC
Amtrium 1+2
Tommy Nguyen, Liftoff Mobile
★★★★☆

Liftoff's custom SRE agent using MCP for automated incident diagnostics — practical but narrowly scoped

Overview

Tommy Nguyen from Liftoff shared their journey building a custom SRE agent powered by MCP. The most interesting aspect was how they integrated the agent directly into their incident management workflow, assisting engineers with diagnostics during incidents.

How It Works

The agent acts as a first responder — when an incident fires, it pulls context from Grafana, Prometheus, Jira, GitHub, and other integrations via MCP and produces a diagnostic summary. The on-call engineer starts with a full picture instead of digging through dashboards manually.

What I Liked

  • Practical, real-world production use case for MCP
  • The agent-in-the-loop approach for incident management is compelling
  • Clear demonstration of MCP’s value as a universal integration layer

What I Missed

The presentation was too focused on a specific scenario. I would have liked to hear more about the journey details:

  • Challenges — what went wrong along the way?
  • Security — how do they handle credentials and access control across all those MCP integrations?
  • Guardrails — what prevents the agent from taking wrong actions or hallucinating diagnostics?
  • Governance — who approves what the agent can access? How do they audit agent actions?
  • HolmesGPT — open-source AI agent for incident investigation, similar concept
  • Liftoff — the mobile growth platform where Tommy’s team built this
← Back to schedule