AuthLN
Book a review

The record Transparency is the product

A record no one can rewrite.

Every sign-in attempt, every grant, every action and every refusal lands on a record the actor cannot rewrite.

Transparency is the product, not a feature. You check the record instead of taking anyone's word for what happened, including ours.

Book an Agent Access ReviewWhat gets recorded
A record held out of reach of the system that actedOn the left, the system that acted: an identity provider, a person and agents. Each event crosses a wall marked out of reach and lands as a write-once block on the right, with its ID: two sign-ins (Authorized, Walked away), a grant, an action and a refusal. Last, an attempt to rewrite the record runs into the wall and is refused. The entries are illustrative.THE SYSTEM THAT ACTEDIdPIdPpersonagentagentrewriteOUT OF REACH8f2c62SIGN-INAuthorized8f2c55SIGN-INWalked awayg-0417GRANTfrom a named persona-0418ACTIONwithin the grantr-0419REFUSALno grant, stoppedREFUSED: WRITE-ONCE
Illustrative entries

Why it sits outside

Written by something other than the thing that acted.

When an agent writes its own account, nothing outside it can verify the accuracy. The record is written once and held where the thing being recorded cannot edit it, so it can be checked.

The system that acted

Apps, agents and your identity provider act here. None of them can edit what lands on the right.

The record · write-once Illustrative entries

  1. 8f2c62Sign-inAuthorizedoutcome, correlation ID, timing
  2. 8f2c55Sign-inWalked awaythe missing device signal is the finding
  3. g-0417Grantauthority, from a named personwhat it covers, and who it was handed toComing soon
  4. a-0418Actionsub-agent 4: update ticketwithin the grant, on the recordComing soon
  5. r-0419Refusalcaller, unverified: request customer recordsno grant, flaggedComing soon
A closed loop that nothing outside the agent checksA circle with four steps, clockwise: the agent acts; the agent writes its own account; we read the account; we call it fine. Nothing outside the loop checks any step.It actsthe agentIt writes the accountthe agentWe read the accountusWe call it fineusA closed loopnothing outside it checks

Without it

An account you cannot check is not evidence.

Today the only account of what an agent did is usually the agent's own log, written by the system that acted. It acts, it writes the account, we read the account, and we call it fine.

The record breaks that loop: it is written outside the system that acted, so anyone who needs to can check it.

What gets recorded

For a sign-in, for a grant, for an agent action.

Factor AssuranceShipped

For a sign-in

  • Outcome + correlation ID. Authorized, Denied or Walked away.
  • Dwell + 4-stage timing
  • IP, ASN and geo-city
  • VPN, Tor and proxy flags, when the network exposes it
  • Device GPS + accuracy, and origin trust class, when they engage

A walk-away still leaves a record. The missing device signal is the finding.

Agent AuthorityComing soon

For a grant

  • The person who gave it. A named person.
  • Who it was handed to, and every hand-off after it, each passing on less
  • Any revocation, which stops every agent acting under that grant

Authority comes from a named person and only narrows as work passes down.

Agent AuthorityComing soon

For an agent action

  • The agent that acted, and whose grant it acted on
  • The action, and what it reached
  • The reason, recorded as the action happens
  • A refusal: the moment it was stopped, and why

The reason is recorded as it happens, not narrated by the agent afterwards.

Who it is for

Your own team, your auditors and your customers.

Your team

Every event arrives resolved, with an outcome and an ID you can act on and join to the logs you already keep.

Your auditors

A durable, correlated decision record, auditable months later, written once and out of reach of the thing it describes.

Your customers

Proof of what happened, and on whose authority, that does not rest on the word of the system that acted.

Start here

Start with what is running today. Book an Agent Access Review.

Two weeks, from read-only exports, with nothing installed. You get an inventory of your agents, a map of what each can reach and under whose authority, and a plan your people own.

Book an Agent Access ReviewWhat the review covers
One person approves one agent. The agent starts sub-agents that fade from view, and the trail ends in a question mark: nobody can say how many there are or who owns them.?

And the questions that follow

  1. 01How many changes can one agent, and the sub-agents it starts, make in our systems?
  2. 02When we revoke a person’s access, do the agents acting for them lose it at the same moment?
  3. 03If an agent is compromised, what did it do, and under whose authority?
  4. 04When did an agent first reach sensitive data, and whose grant let it?