AXES - the Autonomous eXecution Evidence Standard
AXES - the Autonomous eXecution Evidence Standard - is an open, vendor-neutral standard for binding tamper-evident evidence to an autonomous action at the moment it executes, then rendering that one record as role-specific reports a board, auditor, regulator, or engineer can each rely on. Stewarded by Magentix AI. It evidences; it never certifies.
Last updated: 21 October 2026
What AXES is for.
Wherever a consequential action runs without a human at every step - a payment, a manufacturing step, a standard operating procedure, a supply-chain hand-off, a customer or HR decision, an infrastructure change - the same question follows: what was done, under whose authority, and can it be proven later. And each profession needs a different cut of that same truth. A board, an auditor, an operations lead, a disputes team, and a regulator cannot usefully read the same raw stream. They need the same underlying record, framed for the desk they sit at.
AXES is the open standard for binding that evidence at the moment of execution, then rendering it as an Autonomous Execution Report each role could actually use. The situations it covers are not limited to financial services. The breadth of those use cases is wider than any single industry, and wider than any single product.
Two worked examples make that breadth concrete rather than asserted: a financial one, an autonomous accounts-payable run, and an industrial one, an autonomous production batch release. Both are built on the identical evidence skeleton. Finance came first because it sets one of the hardest evidentiary bars in practice, not because it is the boundary.
The same evidence skeleton, two industries. Open either example and you are looking at one record structure - authority, controls, commit, reconciliation, hash chain - doing its job in a different world. A payment instruction and a released machined part are the same shape of evidence. That is what makes AXES a standard rather than a format for one sector.
Captured with the action, not consulted by it.
Most approaches log that a check ran; they do not bind the result to the action in a way that survives policy rotation and later review. A record and a proof are not the same thing.
| A record | A proof (AXES) | |
|---|---|---|
| Captures | that a check ran | the authority state in force |
| Bound to the action | loosely / by reference | by content-addressed fingerprint |
| Survives policy rotation | no | yes - resolves to commit-time state |
| Examinable months later | not reliably | independently, by any party |
How it works.
- Capture once. Execution evidence is bound at the moment of action - authority, artefacts, and outcome - across whatever process the automation ran.
- Render for the role. Board, audit, forensic, regulator, conformance, and machine-readable views are cuts of the same record, not separate stories.
- Keep the report portable. The report is designed to become an organisation's own artefact. Branding belongs to the organisation that runs the process, not to the steward of the standard.
That is why a treasury lead, a fraud analyst, a plant operations manager, and a disputes investigator can open different views and still be looking at the same event.
Open standard. Conforming implementations.
AXES is developed in the open at github.com/magentixai/axes. Schema, documentation, and the Golden Trace evidence bundle are public. Requirements enter a public register that shows its working.
Magentix AI's execution-assurance product, ARBITR, is a conforming implementation of the kind of execution-evidence posture AXES describes. ARBITR is the first conforming implementation, which is how Magentix AI stress-tests the standard against real execution before asking anyone else to adopt it. AXES itself is not ARBITR, and it is not limited to Magentix AI products. Other implementations can conform. The standard is the shared language; the products are how organisations put that language to work.
AXES and x402.
x402 is an open standard for internet-native payments between clients and servers, giving autonomous agents a way to pay for what they use. As agentic payments move from demonstration to production, a question sits immediately downstream of settlement: when a paid-for action is later queried, what record exists of what actually executed, under whose authority, and can a third party verify it without the vendor present. Settlement proves money moved. It does not, on its own, prove what the money was for or that the action behind it was authorised.
That evidence record is where AXES fits. AXES does not adjudicate a payment, reverse it, or enforce a policy. It binds tamper-evident evidence to the action as it executes, so the record can be recomputed from its own bytes and read by whichever role needs it afterwards. The boundary is deliberate, and it is the same boundary AXES holds everywhere: it evidences; it never certifies.
Magentix AI takes this to the standard body directly. Martin Sansone has put the post-settlement accountability scope question to the x402 Foundation technical governance and proposed chartering an evidence-record working group, scoped to the record shape and a shared conformance corpus rather than to adjudication or enforcement. The AXES Golden Trace is offered there as one worked conformance corpus. The intent is to keep AXES aligned with x402 as the ecosystem forms, so that evidence produced under one standard reads cleanly under the other.
On the roadmap: an open AXES SDK.
AXES is a standard first. A standard is most useful when anyone can produce and check evidence against it without asking permission. An open AXES SDK is in development to make that practical: libraries to emit Standards Envelopes from a running process, and to validate a bundle locally against the conformance ladder before it leaves the building.
The same SDK is the route to evaluate ARBITR independently. Emit a conforming bundle, submit it to an ARBITR evaluation endpoint, and receive the rendered Autonomous Execution Reports back, without a pilot agreement in place. Conformance stays open and checkable by anyone; the rendering is how Magentix AI shows what a conforming implementation does with the record.
This is on the roadmap and not yet available. It will be published at github.com/magentixai/axes when it is ready. To hear when it lands, follow the standard below.
Read an example. Shape the next draft.
The AXES Example publishes two worked evidence events, one financial and one industrial, each as several role-specific Autonomous Execution Reports. Choose the view that matches your world, read it as a report that crossed your desk, and answer what would need to change before you could rely on it. Your answers enter the public requirements register.