Sgovr
Blog
Practical guides, technical deep-dives and industry analysis on AI agent infrastructure - discovery, mandates, payments, trust and the open protocols that make agents safe to deploy.
Capability specs as contracts: how agents reason about risk before calling
OpenAPI tells developers what an API does. A capability spec tells an agent what an action costs, what authority it requires, and what happens if something goes wrong - before the call is made.
The three layers every agentic commerce platform needs
Discovery, Trust, Settlement - and why you can't bolt trust onto an existing REST API after the fact.
Agent identity is not user identity
OAuth scopes were designed for users. Service accounts were designed for services. Neither maps cleanly onto an AI agent acting on behalf of a human principal it may never meet. Here is what a mandate-based identity model looks like and why the distinction matters.
What happens when your AI agent double-books a hotel room?
A network timeout. A retry. Two reservations, two charges, one very confused guest. Here is the failure mode most AI booking agents are one bad connection away from - and the infrastructure that prevents it.
Why API keys are the wrong auth primitive for AI agents
You wired your agent to a service with a shared API key and now something feels off. That instinct is correct - keys authenticate callers, but agents act on behalf of human principals with scoped, time-limited authority. Here is what mandates give you that keys never can.
MCP gives agents tools. What gives agents trust?
MCP standardises how agents call tools - but it deliberately says nothing about authorisation, spend caps, payment, or audit. Here is what fills that gap and why it matters.
How AI agents can book hotels without scraping
Web scraping is brittle, slow, and legally grey. Here is how purpose-built agent APIs make hotel booking - and every transactional workflow - reliable, safe, and auditable for AI agents.