Databricks engineer's playbook: MCP as a governed data layer with per-user perms and audit
wesleydias84 · reddit · 2026-10-10
A Databricks engineer (disclosed) argues MCP is more useful to enterprises as a governed data layer than as a way to hand agents tools: Claude/Codex/Cursor read real business data while entitlements and audit stay intact.
Key design points from his hands-on test via a governed gateway + MCP servers:
- The data MCP is fail-closed and won't start without a valid identity
- Reads are scoped to the caller's catalog permissions — the agent only sees what the user can
- Every model and tool call is attributable per user
- Approval/sandbox gates on any writes
He ends with an open question: pass caller identity through the MCP layer, or run servers as a service principal? Full write-up linked.
More from coding & agent
- Claude cut a 2.6-hour video export to 20 minutes — but only after being told to rebuild the server — byhewar · 2026-10-23
- 3 flight-booking rules my AI agent checks that I'd never do by hand — Efistoffeles · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Building an autonomous SEO agent in Hermes with three skills: programmatic pages, Keywords Everywhere, DataForSEO — EXM7777 · 2026-10-23
- Building a sports betting AI agent with onchain execution and human approval by default: architecture and lessons — Dry-Comfortable-2678 · 2026-10-23
- How do you audit AI agent decisions in regulated workflows? A field survey — Ok_Way_1993 · 2026-10-23