— INTREXA R&D
The questions behind the next product.
As AI agents move from answering to acting, new questions appear about authority, validity and trust. We study them in the open, and build products where the answers become software.
— Research areas
Four questions we keep coming back to.
Two are active and already shape the specifications and AXIS. Two are directions we are exploring. Select one to see the question, where it lives today, and what is still open.
THE QUESTION
How should an agent's identity and delegated authority be represented so any party can check them?
Authentication says who is asking. It says nothing about what that agent was actually authorized to do, or on whose behalf.
IN AGF (OPEN)
Spec 01 · Delegation token, Spec 03 · Trust zones
IN AXIS (PRODUCT)
Evaluates delegation chains and trust before every decision.
OPEN QUESTIONS
- ?How should authority narrow at each delegation hop?
- ?What should a verifier be able to prove without calling home?
READ MORE
— From research to product
Research that ends in something you can run.
01
A question
A concrete problem teams hit when agents start acting.
02
An open model
Specifications and RFCs anyone can read, challenge and implement.
03
A working product
AXIS implements the model and tests it against real usage.
04
What we learned
Findings, caveats and corrections published as writing.
Example: the AGF execution-time validation specification (Spec 30) came first; AXIS implements it in its gateways; the caveats we found are in our published writing.
— Get involved
Follow the work, challenge it, or help shape it.
The specifications have an open RFC process. Working on similar questions? We would like to hear from you.