Uber Scales MCP to 60,000 AI Agent Tasks/Week with Claude/Cursor Harnesses
Key Questions
How has Uber scaled AI agent tasks?
Uber scales to 60,000 AI agent tasks per week using MCP with Claude and Cursor harnesses. 90% of work for 5,000 engineers is AI-driven via Minions.
What is MCP and why does Uber prefer it over CLI?
MCP standardizes agent-tool interactions, preferred by Uber for governance over raw APIs or CLI. It supports sub-agent patterns and multi-agent UIs efficiently.
How does n8n integrate with Uber's AI workflow?
n8n provides human-in-the-loop automation and new tools that supercharge Claude Code. It reinforces enterprise governance in Uber's MCP scaling.
5k engs 90% AI Minions 1.8k changes/wk; governance OpenAI/MS/JetBrains/Cursor/n8n; MCP std over APIs; sub-agents/multi-UIs; hybrids.
Sources (2)
Updated May 5, 2026