Identity for AI Agents
summary
Okta-backed stack adds a token vault that persists refresh tokens, manages lifetimes and does token exchange so agents get upstream API access without reauthenticating; an async approval flow uses client‑initiated back‑channel auth (CIBA-style) to send structured approval requests and return scoped access tokens; an MCP server is modeled as a client with dynamic client registration (DCR) support; a Next.js agent + MCP demo deployed on Vercel shows the flows working end‑to‑end (portfolio read, gated create‑order approval).
AI-written summary · about
chapters (15)
- 0:00Start
- 0:14AI agent identity
- 3:09Vision and challenges
- 8:59AI approval feature
- 11:05Token vault for access
- 15:23MCP servers as clients
- 20:45Downstream chatbot with identity
- 27:58Account connection and token exchange
- 34:04Agent access to resources
- 39:52MCP server permissions
- 44:27Dynamic client registration
- 55:21Async authorization
- 1:08:52Async authorization features
- 1:11:05AI platform integrations
- 1:20:26Q&A
related talks
25mWho Let the Agents Out? Securing AI Workflows the Right Way
17mDon't give your agent the master key
19mSecuring Agents with Open Standards
29mHow Auth0 Secures AI Authentication & Authorization for AI Agents
35mIdentity and Access Management for the Decade Ahead
23mIdentity Best Practices to Secure & Scale Agentic AI / MCP Deployments
23mWho is Calling? Identity for Agents with AAuth
40mIn AI We Trust? Securing the Future, One Agent at a Time