Google Cloud Rapid Agent

Gemini Operations Navigator

For support operations managers handling refund queues: Gemini drafts from policy evidence, MCP tools verify context, Phoenix MCP exposes trace tooling, and human approval blocks risky customer-facing action.

Phoenix MCP verified Agent Builder manifest Vertex rerun status OpenInference trace

Open Workflow Terminal Run YouTube Demo Architecture Shared Timeline Submission Package GitHub

What The Demo Shows

Target user: support operations managers responsible for refund and escalation queues.
Problem: AI support work becomes unsafe when evidence, tool calls, cost, and send approval are hidden.
AI use: Gemini drafts from evidence, MCP tools retrieve context, Phoenix MCP exposes observability, and approval blocks the final customer send.

Gemini Operations Navigator architecture diagram

Verify Locally

bash rapid-agent/scripts/run_google_local_checks.sh