AI Agent Control Checklists
Tiered controls for identity, secrets and egress, audit-event completeness, evidence portability, and exit — with the inspection that satisfies each one.
Published method
Our method, published in full. Judge it by reading it.
We do not ask you to take our word for how we work. These are the frameworks we actually apply to production AI systems, published whole and free to read. Nothing here is a case study and nothing claims a client outcome — the documents are the evidence, and you are the judge.
The direct answer
Why publish the method instead of case studies?
Because a case study asks you to trust a story about someone else, and a framework lets you check the work yourself.
Most AI engineering pages ask you to believe a result you cannot verify: a logo, a percentage, a testimonial from a company you cannot call. We publish the price, the method, and the pass/fail criteria instead. If the thinking below is good, that is a fact you can establish in ten minutes without speaking to us. If it is not, you have lost nothing.
The frameworks
AI Agent Security Threat Framework
The threats a production agent is exposed to, the control that answers each one, and the evidence that demonstrates the control actually holds.
Agent Reference Architectures
Architecture patterns for production agents, chosen from constraints and tested behaviour rather than from a diagram or a vendor feature list.
Enterprise Agent Runtime Landscape
What each enterprise agent runtime documents it supports, recorded as vendor-documented scope rather than as a recommendation.
AI and Agent Governance Framework
Who may approve, change, operate, suspend and retire an agent — through which lifecycle gates, with which registry record and which retained evidence.
Walkthroughs
Recorded walkthroughs
Short recorded walkthroughs of how each framework is applied to a real production decision. These are in production now; the slots below are reserved and will fill in place.
Coming soon
Coming soon
Coming soon
What this is not
What these documents do not claim
These frameworks demonstrate method. They are not proof of an outcome, and we will not present them as one.
- No client is named in any published document, and none contains a client outcome, testimonial, or case study.
- No result is implied. A framework shows how a decision gets made, not that a particular decision worked out.
- External claims carry a citation marker into that document's source register. The library's own rule is that an unregistered claim does not pass review.
Next decision
Apply the framework to your work
The frameworks and their document downloads are free, with no email required. If you want help applying the method to your system, leave an email for an engineering conversation.
We review your enquiry and reply by email. No commitment to a paid project; scope and terms are agreed before work begins.
Prefer email? dev@alpha-ai.dev