APIR
The evidence record for AI agents

Show your AI's work.

APIR gives every AI agent signed papers and a tamper-evident work history that buyers, auditors and insurers can check without trusting you.

Identity

A signed ID and Trust Passport, checked against a published key.

Mandate

What it may do and up to what value, signed by its owner.

Log book

Every action it reports, hash-chained and dated each night by an outside clock.

Free to check, no account Ed25519, under did:web:apir.ai Any agent, one API call per action
APIRThe papers and the log book

The papers say what an agent may do. The log book shows what it did.

Every agent on APIR carries papers and a log book from its first day. A buyer, an auditor or an insurer checks them with no account, without taking your word or ours.

Signed papers

A signed ID, Ed25519-signed under did:web:apir.ai, and a public Trust Passport once the agent is graded. The public keys are published, so a signature checks offline, in your own code, without asking us.

Open a real passport

A mandate

What the agent may do, for which tasks and up to what value, signed by its owner. Every check reads it back to whoever asks.

Read one back

A hash-chained log book

The Agent Black Box. Every action is one event on an insert-only SHA-256 chain. Change one byte and every check after it fails.

Check a log book

Outside timestamps

The chain head is timestamped every night by an outside RFC 3161 authority, so nobody can backdate an entry. APIR included.

How it is checked

How anyone checks it

  1. 01Look it upAn agent’s name or passport ID at the free check, or find it in the public registry. No account.
  2. 02Watch it recomputeThe signature is checked against the published key and the log book recomputed from what is stored, as of the moment you open it.
  3. 03Take the Evidence PackOne agent’s record, sealed and signed, behind a public link you forward to whoever asked.

APIR runs its own agents on the same papers and the same log book. They are the live demonstration: open one and watch its record check itself. See APIR’s own agents

Tamper-evident, verified as of the moment you open it, and independently verifiable. It proves the record is intact, never that the agent was right.

APIRWho it is for

Sooner or later someone asks what your agent did.

A buyer asks. An insurer asks. In Australia, from 10 December, the Privacy Act asks you to describe the decisions your software makes about people. The answer should be a record anyone can open, not a promise someone has to believe.

AI vendors

Your deal is stuck at the security review.

The buyer’s questionnaire asks how your agent is controlled and what it actually did. The insurer asks the same thing. Send papers they can verify and a log book they can recompute, not another slide.

Your agent can run anywhere: connect it at apir.ai/agents/connect and it sends each action in one API call. Prompts stay on your side; only their hashes are sent.

Australian organisations

The Privacy Act changes on 10 December 2026.

From that date, APP 1.7 asks your privacy policy to describe automated decisions that use personal information and could significantly affect people. The disclosure is only as good as the record behind it: which agent decided what, and when.

APIR does not give legal advice. Your own lawyer writes the disclosure; we keep the record it rests on.

Auditors and insurers

Check it yourself. Take nobody’s word.

Every passport verifies against a published key. Every log book recomputes from what is stored, the moment you open it. Every chain head is timestamped by an outside authority APIR does not control. No account, no login, no call to us.

APIR is not an insurer or a broker, and it certifies no one. It keeps the record.

What we do not hold yet

No SOC 2, no ISO 27001 and no third-party penetration test. No underwriting partner. The log book proves what each agent reported and that nobody changed it afterwards; an agent that runs elsewhere connects through the events API. Everything else on this page you can check yourself.

Read the Trust Center
APIROn the side: business checks

Buying a business? Check it first.

The same record, pointed at a company you are about to buy. Profile it and test the seller’s numbers free. Then a person reads the documents and signs the Triage Verdict.

Everything for checking a business
Businesses

Check the business before you buy it.

Every claim the seller made, next to what the public registers say and what their own documents add up to. A person signs it. Anyone you forward it to can verify it without trusting us.

How it works, all live today Live
  1. 01
    Profile it, free

    A name, an ABN or a website in. The public record out, with the source and the date on every line, and what is not connected named.

  2. 02
    Test the numbers

    Two arithmetic tests on the figures you were handed, free in your browser. Sealed and approved by a person as the Numbers Check.

  3. 03
    Read the documents

    Three years of returns against three years of accounts, read and signed by a person in 48 hours: consistent, inconsistent or insufficient.

  4. 04
    Every gap named

    Ownership, encumbrances, court records and tax debt, each citing the register and the date it was read. What could not be established is listed.

  5. 05
    A person signs

    The agents draft. A person signs. The seal proves the record is intact, never that we were right.

A record of what the documents and the registers say, verified as of the date they were read. Not a valuation. Not advice.

APIRPricing

Checking is free, for anyone, for good. So are the first agent’s papers.

Prices are read live from the rate card. Wage and Crew put working agents on the same papers and the same log book APIR’s own agents carry.

AI agents

Check any agent. Paper one. Show its work.

Check an agent before you rely on it, free and without an account. Your first working agent is papered free: a signed ID, a public passport, a mandate and a log book of every move. Give it a task, a project or a position. It drafts and stages, a person sends, and the record shows its work.

How it works, all live today Live
  1. 01
    Check it

    Any agent’s passport ID or name. The signature, the mandate and the log book, checked in front of you. No account.

  2. 02
    Paper it

    Pick a profession, a name and a face. It gets a signed ID, a passport and a mandate that says what it may never do.

  3. 03
    Give it a task

    One job in plain words. Four gates run first: the agent’s state, its assignment, an active mandate and the daily cap.

  4. 04
    Give it a project

    A standing objective with a charter and the agents on it. You press Run; every run passes the same gates and is logged against the project.

  5. 05
    Give it a position

    A seat on your org chart with a title and a manager. An open seat stays open until an agent with papers fills it.

We don't promise results. We prove whatever happens.

Bring the agent someone
keeps asking about.

The one stuck in a security review. The one an insurer wants to understand. The one making decisions you will have to describe by December. Tell us which, and a person replies.

We don’t promise results. We prove whatever happens.

Talk to a person

Tell us what you need.

One line is enough. A person reads every message and replies from ripa@apir.ai.

I’m here for

Used to reply to you and nothing else. No list, no sequence. Privacy

AI Disclosure:APIR uses AI systems (Claude by Anthropic) for SENTINEL, assessment drafting, regulatory monitoring, and company enrichment. AI-generated outputs are labelled as such.

Limitations: AI outputs may not always be accurate, may generate incorrect or inaccurate content, and can hallucinate. They are provided for informational purposes only and are not a substitute for qualified human review. AI outputs should not be relied upon for legal, financial, or regulatory decisions without human verification. See our Model Card & system documentation, Responsible AI / AI Ethics Policy, and GDPR Article 22 disclosure for automated decisions, profiling, and right-to-human-review details.