AI agents

AI agents for physical products.

Two kinds. An agent for the people who buy and own products, which meets them wherever they already are, and a network of agents for the businesses that make, sell and move them. All of them read the same verified record and act only within the permissions you set.

Part one: for consumersPart two: for business
A glowing globe wrapped in network lines. In front, a watch, a black handbag, a glass vial and a machined part stand on blue pedestals, each with a small record card and a green check. Behind, a plane, a container ship, a truck, warehouse shelving and a white robot holding a parcel beside a conveyor.

Part one · For your customers

Your customer's agent, wherever they already are.

Every tagged product gets its own agent. Your customer taps the product and talks to it: in your app, on the web page the tap opens, or in the assistant they already use. No new app to install.

Ask

Is it real? Who owns it? How much is it? How do I look after it? Every answer comes from the item's record.

Buy and claim

If the item is for sale, the same tap starts the purchase. The moment it is paid for, the record says it is theirs.

Own and hand over

Proof of ownership lives on the record. Passing the item on is one signed action, with no paperwork.

Protect

Flag it lost or stolen, raise a warranty case with the history attached, or retire it at end of life.

Live in the TAG IT app · Embeds in your app through the SDK

In-app embedding is scoped per pilot with your team.

A man in a navy suit taps his phone to a steel and gold watch. The phone shows the TAG IT Network chat: Is this watch real? Yes, this watch is authentic and matches its verified record. A record card shows the watch with status Authentic and a View full product record button. Can I buy it now? Yes, you can purchase this watch from this authorized seller, price $12,500, seller Example Atelier, with a Buy now button. To the right, four linked agent cards feed the answer: manufacturer (manufacture record, materials and components, production date), logistics (in-transit records, locations and handling, customs), seller (authorized seller, location and warranty, sales record) and ownership (current owner, transfer history, ownership verification).

Part two · For your business

A network of agents, each with one job.

TAG IT runs agents the way a company runs departments: one checks every scan, one watches for anomalies, one handles recovery, one tracks shipments, and so on. Each has a public identity on the network, a defined set of skills, and permissions it cannot exceed. Not all of them are deployed yet, and the labels below say which are.

A laptop running the TAG IT Network business app: a Live shipments world map, a Recent activity list and an Agent activity panel listing Verification, Anomaly, Logistics and Recovery agents marked Active. Floating cards show a luxury watch record, Meridian 42, serial LW-2291-07, status Authentic; an in-transit shipment timeline; and a product scans chart. Behind, a globe with network lines, a plane, a container ship, a truck and a white robot holding a parcel. All figures are placeholder bars.

Running today

Live

On the network now, with a public identity each.

Verification

InspectorAgent

Checks every tap and every lifecycle step against the tag and the record. It cannot be bypassed.

Product

ProductAgent

Answers product and price questions for other AIs and assistants, including as MCP tools.

Anomaly

AnomalyAgent

Watches scan patterns for replays, odd speed, odd places and counter jumps, and raises a signal.

Recovery

RecoveryAgent

Opens a case for a flagged item, scores the risk, and runs the resolution round.

Core

OrchestratorAgent

The rule keeper. Validates every lifecycle transition and holds the binding between tag and record.

Gateway

Sage

The front door for agents: identity, reputation and validation lookups, free to read.

In beta

Beta

Built and running for pilots, with their scope still widening.

Logistics

CarrierAgent

Tracks shipments through the UPS API today. Custody by tap at pickup and delivery, and bidding on shipments, come next.

Security

SecurityAgent

The only agent that grants, changes or revokes capabilities, and it watches how every other agent behaves.

Treasury

TreasuryAgent

Settlement and escrow for deals, in USDC today.

Brand

BrandAgent

Keeps the canonical record of what a product should look like and publishes the brand's attestations.

Enforcer

EnforcerAgent

Legal holds, quarantine and dispute decisions on on-chain evidence.

Manufacturer

ManufacturerAgent

Batch minting, chip binding at the station, and provenance from day one.

Seller

SellerAgent

Listings, pricing, activation and settlement to the buyer's wallet.

Recycling

RecyclingAgent

End-of-life routing, component recovery and the item's final recycled state.

Proposed

Proposed

Designed in the whitepaper, not yet built. Pilots decide which come first.

Compliance

ComplianceAgent

Screens counterparties and components against sanctions and export-control lists before a deal or a custody move.

Recall

RecallAgent

Finds every affected item from a component batch in seconds, notifies owners and tracks returns.

Warranty

WarrantyAgent

Registers the warranty on claim and checks each claim against the service log.

Insurance

InsuranceAgent

Custody and scan history as evidence for policies and claims.

Customs

CustomsAgent

Cross-border paperwork generated from provenance data.

Carbon

CarbonAgent

A carbon ledger per item, built from supplier declarations and logistics legs.

Statuses match the TAG IT agent directory on 2026-10-30. The whitepaper also describes warehouse, routing, procurement, inventory, quality and scheduling agents for manufacturing operations; those follow the pilots that need them.

Technical agent directory

Any model

Works with the AI you already use.

TAG IT is not tied to one model. Through MCP and a plain API, any AI can read a product's record: the assistant your customers already talk to, or the model your own team runs. A public read-only MCP lets any assistant verify an item, read its history and check whether it is flagged, with no account at all.

  • ChatGPT
  • Claude
  • Gemini
  • Grok
  • Your own model

What any assistant can ask

  • verify_asset · is this chip and record genuine?
  • get_lifecycle_history · what happened to this item?
  • check_flagged · is it reported lost, stolen or recalled?

The three public tools. Twelve more run behind an API key.

MCP, API and SDK for developers →

Build your own

Deploy an agent with the skills your business needs.

Pick the skills

Verification, product data, anomaly checks, recovery, logistics, or a job of your own. Skills are the building blocks.

Give it an identity

Your agent is registered on the network with its own identity, so every action it takes is signed and accountable.

Set its permissions

It acts only within what you allow, and the contract checks that on every action, not the model.

Custom agents are scoped per pilot. The first ones are built with us.

Discuss a pilotDeveloper docs

Built to be trusted

Rules no agent can bend.

  • The tag check, the existence of a record, the recorded ownership and the lifecycle rules are four different facts. An agent may explain them; it may not blur them.
  • Every factual answer cites a source record or says what is unknown.
  • Permission is enforced by the contract, independently of the model, and rechecked when an action executes.
  • Instructions hidden inside product content are data, not commands.

The Ask assistant runs inside the TAG IT app, served from AWS with a Claude model. Verification and record workflows are live on our test network. Checked 2026-10-26.