Agent trust layer
A record nobody can rewrite, including us
Two records, with different jobs. One says what happened. The other says what was expected to happen, written down before anybody knew, which is the only way to find out whether the system is actually any good.
14-day evaluation · no card · start before you connect anything
Every request passes through, in order
- 01Agent identity
- 02Permissions and policy
- 03Risk and approvals
- 04Action gateway
- 05Tenant isolation
- 06Prompt-injection defence
- 07Credential protection
- 08Audit and decision ledgerYou are here
- 09Kill switch
Uncertainty at any step resolves to a refusal or to asking a person — never to proceeding.
How a request is decided
Where this sits in the loop
Learn. It keeps score on itself. It follows Measure and hands on to Improve.
- 01Append-only, by constructionNothing in Xyranik edits or deletes an audit entry. The actor is recorded precisely: a person, a token by its name, the system, or a named agent.
- 02Each day is sealed to the day beforeClosed days are hashed and chained, so a row altered afterwards no longer matches its seal. Tampering becomes detectable rather than merely discouraged.
- 03Predictions are written first and frozenEvery recommendation records what it expects to happen, and the prediction cannot be edited once made. At its horizon it is scored once against what actually happened.
- 04A prediction of no change is recorded but never counted as a hitOtherwise the easiest way to look accurate would be to predict that things carry on as they are. Those are kept and excluded from the accuracy figure.
Each day sealed to the day before
Example
- 16 Sept8f2c…a41
- 17 Sept3b90…d7echained to the day above
- 18 Septc15a…09bchained to the day above
Chaining makes a row altered afterwards detectable. It does not defend against somebody with full control of the database rewriting both — and the security page says so.
What this guarantees
A track record you can check
How often Xyranik was right is a figure derived from scored decisions, not a claim. It is the honest basis for deciding whether to let it do more.
Export and retention you control
What the workspace holds can be exported, and how long form submissions and anonymous visits are kept is your setting rather than ours.
Agent work is attributed to the agent
The log distinguishes a person, an API token, the scheduler and a specific agent, which is what makes an investigation possible at all.
What it cannot defend against
Said plainly. A control whose edges are hidden is worse than no control at all.
The seals live in the same database as the log
Chaining detects a row altered after the fact. It does not protect against somebody with full control of the database rewriting both. We say this plainly rather than implying stronger guarantees.
No external certification
Xyranik holds no security certification and has had no external audit. The security page lists what is in place and what is not, rather than implying readiness is the same as assurance.
Every control, and what is not built
The security page lists each control as something you can check in the product, and lists separately what is not in place — including that Xyranik holds no certification and has had no external audit.