Agentic Development at Enterprise Scale: RAG, MCP & Guardrails
summary
Enterprise solution shows a concrete architecture that uses RAG to ground LLMs on approved enterprise docs (10M+ chunks, top‑K=15) combined with MCP (Model Context Protocol) as a tool/resource/prompt broker where MCP client/server each re‑authenticate before calling Jira or other APIs. Proposed agent actions hit a policy/guardrail layer (risk, policy, privilege, impact, blast radius) that yields allow / human‑approval / reject, logs to an audit trail, and enforces least privilege and staged autonomy (L0→auto‑merge).
AI-written summary · about
chapters (8)
- 0:00Agentic development at scale
- 3:14Enterprise AI adoption challenges
- 7:53Regulated agentic systems architecture
- 10:42RAG for secure enterprise grounding
- 15:15MCP as enterprise capability layer
- 20:10Guardrail decision making
- 24:27Autonomy levels in agents
- 26:59Measuring outcomes & governance
related talks
23mZero-Trust Architecture for Agentic AI: Securing Multi-User Access and Third-Party Integrations
47mMCP Demystified: Cutting Through the Hype-What It Is, What Isn't & Its Role in the Agentic RAG Stack
26mArchitecting Enterprise Agents Ecosystem: Standardizing APIs with MCP and Orch…
29mMCP Security: Keep Your AI Agents from Spilling the Tea
18mThe "Agentic Organisation": Governing Multi-Agent Developer Teams with Nutanix Agent Gateway and MCP
20mBuilding intelligent, secure, and scalable autonomous systems
22mAgentic AI for Engineering Leaders | Strategy & Execution Guide
17mOur Journey to Safely Unleash Agents at Meta Scale