Notes on agent identity and governance.
Written plainly, for the people who have to decide whether to let an agent touch production: what a verifiable identity actually buys you, what it does not, and what is worth asking before you take either on trust.
Identity is half the problem
A verified identity answers who acted. It does not answer what they should have been allowed to do. Here is why agent permission has to be decided per tool action, not per agent.
Read the article → 21 October 2026Can AI agent governance run in your own cloud?
The enterprise deployment question, answered plainly: what runs today, what is being built, and what has not started, so a security or infrastructure review can plan around the real state rather than the pitch.
Read the article → 22 October 2026What VAID is, and why delegation can only narrow
For engineers: what a VAID actually is, the exact bytes that get signed, and the mechanism that makes delegation a one-way narrowing rather than a trust assumption. Apache-2.0, with the repo linked.
Read the article → 23 October 2026Revocation that tells the truth
Nobody's revocation is instant. Here is what is true: a revoked credential is refused the next time it is checked, and the exposure is bounded by how long that credential was ever good for.
Read the article → 24 October 2026Ten questions a risk team should ask before agents touch production
A working checklist, not a sales pitch: ten specific questions a risk or security team can put to any vendor, including us, before letting an agent touch anything that matters.
Read the article →