Securing AI Agents in Practice
A working proof of concept: agent-scoped OBO tokens, RFC 8693 token exchange, and delegation chains that show exactly which agent acted for which user.
Explorations in software engineering, technical writing, and minimalist design.
A working proof of concept: agent-scoped OBO tokens, RFC 8693 token exchange, and delegation chains that show exactly which agent acted for which user.
Give each AI agent its own OAuth 2.0 identity so permissions are enforced at the API layer and every action is auditable, illustrated with a corporate name change scenario in Singapore banking.