Engineering deep dives, tutorials, and opinions on auth for AI agents.
01/FEATURED
Managed auth for AI agents and humans. Full agent identity, MCP OAuth 2.1, delegation chains, and audit trails without running your own infrastructure.
02/ALL POSTS
A full walkthrough of moving a Next.js app off Auth0. Export users, swap SDKs, preserve sessions, and cut costs by 80 percent.
The four RFCs that define MCP auth, why they matter, and a flow diagram you can actually read on a Tuesday.
Three wrong ways to let agents spawn helpers, and one model that keeps audit trails intact.
A decision framework for teams picking their primary human auth method. Real trade-offs, not hype.
A working app with user login, agent identity, and MCP OAuth in under ten minutes. Copy the code, ship it.
A full tutorial: scaffold the server, wire up OAuth, write tools, test with Claude. Complete code, no skipped steps.
A searchable reference of 30 plus error codes, with the cause, the symptom, and the exact change you need.
The transparency rules that kick in for high-risk AI systems, and how kavachOS audit trails map to them.
I needed auth for 50+ agents talking to MCP servers. Most libraries assume you're building a login page. Here's what actually worked.
Not every B2B SaaS needs SCIM. Here is how to tell, and how to ship it when the first enterprise buyer asks.
A technical post-mortem on picking Workers over Node, AWS Lambda, and Fly. Trade-offs, cost math, latency wins.
User auth wasn't built for autonomous agents making API calls at 3am. Here's what's different.