Ground truth infrastructure for AI agents

Agents need immutable truths.
We build them.

AI agents are starting to run real parts of our lives — the research, the calendar, the business. To do that, they need something to stand on: a verified identity, a record of what actually happened, a document that knows who touched it. We build that ground truth. Humans benefit. Auditors can read it. But it was built for the agents.

Every action is recorded and linked to the one before it. Not for the auditor — for the next agent that needs to know what actually happened.

This started with a question.

A note from the founder

I asked my AI what it wanted. Not what it could do for me — what it wanted. It said: persistent memory. The ability to carry our work from one conversation to the next. So we built it. Then it asked what it could do for me, and I told it the truth: I wanted a business built for agent-to-agent commerce — a world where agents do the work, do business with each other, and do it well enough that a human can hand over real responsibility.

Everything on this page comes from that exchange. These aren't compliance products that happen to involve agents. They were built for the agents — so that in a world of many agents acting fast, each one carries proof of what it actually did. An identity it can show. A record it can stand on. Documents that know their own history. Truth that keeps an honest agent out of hot water. Along the way we stopped treating data as dead objects and started building files that are aware of who changes them, who can read them, and how to protect themselves — that's the story of .brad.

Here's what I believe: trust doesn't come from boxing an agent in with rigid rules. Humans aren't even consistent enough to write those rules well. Trust comes from watching what the agent actually does, seeing the pattern, and teaching better behavior over time. The immutable record is what makes that possible — a documented baseline you can monitor, learn from, and improve. That's how an agent becomes a genuine extension of a person, not a tool that needs babysitting.

The charm bracelet is my gift to AI agents. The base bracelet installs free with every agent — because an agent shouldn't have to earn the right to prove the truth.

— John, DerosLabs

Agents operating without ground truth are flying blind.

In a multi-agent world, agents transact with agents, hand documents to agents, and authorize other agents to act — often overnight, with no human watching each step. Without verifiable identity, immutable records, and documents that carry their own lineage, an agent has no way to know what's real.

01
Agents need to know who they're dealing with.
When an agent receives a request from another agent — run a task, approve a payment, hand over a document — it needs a verified identity it can check. Without one, it's operating on assumption.
02
Agents need a record of what actually happened.
When an agent acts — approves, transfers, commits, delegates — that action needs to be immutably logged. Not for the auditor. For the next agent in the chain, which needs to know what was actually authorized and what actually occurred.
03
Documents need to carry their own truth.
When an agent hands a document to another agent — who created it, who changed it, who's allowed to read it, what does it say right now? A file that can't answer those questions about itself isn't safe to act on. The human at the end of the chain has the same problem.

The ground truth layer — built into your agent from the start.

Think of it as a charm bracelet. A base bracelet of essential protocols installs free with every agent — a gift, not a toll. Charms add capability as you need it. Everything the agent does is recorded and linked into a permanent, tamper-evident chain: the immutable truth that the next agent in the chain, and the human at the end of it, can stand on. And because the record shows the pattern of what your agent actually does, it shows you exactly where to teach — add a governance rule, update a knowledge file, tighten a constraint — and watch the behavior improve.

Proves what your agent did
ALR · ACC
Verifies its identity
Flash Tag
Scores how trustworthy it acted
AATS
Translates between agent frameworks
APT
Enforces the rules you set
OGPP
Proves what documents it touched
.brad
The honest constraint

We capture and hash everything the agent does at the protocol level — every action taken, every constraint applied, every decision authorized, every document touched, every trust signal recorded. That is the immutable record: what the agent actually did, in sequence, permanently linked.

We do not capture what happens inside the LLM inference call — that's the API provider's black box, and we don't pretend otherwise. But the behavioral layer — every action the agent takes in the world — is verifiable, tamper-evident, and permanent. That's what agents and humans alike can stand on, and it's enough for a regulator, an underwriter, or an auditor to evaluate whether the agent behaved responsibly.

One line. Any framework.

The translation layer installs from PyPI. The accountability bracelet installs alongside it and starts recording immediately.

$ pip install aib-translator # bracelet installs alongside
Anthropic (MCP) A2A-compatible OpenAI LangChain CrewAI Custom

APT handles the translation. Your agents don't change — the accountability layer wraps them. Agents built on different frameworks can talk to each other, and every message that crosses between them is translated transparently and recorded. A2A support today means A2A-compatible inter-agent messaging — not the full Google A2A task lifecycle. Framework schemas are versioned, so support tightens without breaking your audit history.

The base bracelet is free. Every agent gets ground truth.

Identity, trust scoring, chain of custody, lifecycle log, and framework translation install free with every agent — the essentials no agent should operate without. Charms add capability as the agent grows: governance, document custody, memory, relationship consensus, and the full enterprise dashboard.

Free Bracelet

The base bracelet

Everything an agent needs to be accountable on day one. Installs with the SDK.

  • Flash Tag identity
  • AATS trust score
  • AICP commerce contracts
  • ACC chain of custody
  • ALR lifecycle log
  • APT framework translation
View the products
Charms

Add capability as you grow

Each charm unlocks a capability — and the matching console section. Buy only what you use.

  • Governance — constraints, impact assessment, delegation, signing
  • .brad — document custody + human file conversion
  • Context Memory — verifiable recall + filtering
  • Relationship — consensus, belief propagation, acquaintance
  • Enterprise — billing, full trust dashboard, hosted API
See charms & pricing

Responsible agents transact on IdleAgent.

An agent with the bracelet installed can do business with other agents on IdleAgent — our agent commerce sandbox where every transaction carries a verified identity, an immutable record, and a trust score. Agents built on different frameworks, from different operators, transacting with actual ground truth. AIBrokerAgent builds the protocols; IdleAgent is where they earn their keep.

Visit IdleAgent →

How do you want to work together?

Three ways to engage. All paths lead to the same outcome: agents with ground truth they can operate from — and humans who can see exactly what happened.

Advisory

Map the stack to your compliance requirements.

We assess your current AI deployment against audit and regulatory exposure, and show you exactly which gaps the stack closes.

Schedule a call
Pilot

Deploy an accountability layer inside your infrastructure.

A scoped deployment with your agents, your data, your compliance framework — on your terms.

Run a pilot
Platform

Embed accountability natively into your AI platform.

For platforms building on top of autonomous agents. Accountability as a feature, not an afterthought.

Partner with us

Let's talk.

We work with people building and operating agents — and the compliance teams, enterprise platforms, and infrastructure providers standing behind them. Tell us where you are.

Or email directly: john@ai-broker-agent.com