Who is acting?
A technical identifier does not always explain who operates the agent or what its purpose is.
Verify the context before exposing a sensitive action.How we evaluate contextShopping with AI assistants
Sell to AI agents on your rules.
Decide what an assistant can do and when it needs approval. Test the rules your platform supports in a guided pilot. Review the available evidence for the operation.
Free analysis: no account needed to start. Pilot: we agree on the scope with your team.
Purchase example · in pilot
Delegated requestBuy 2 units · EUR 124.00
checkout01
Agent identitySignature and context evaluated
to verify02
Policy and riskThe amount requires confirmation
review03
Merchant statePrice and stock checked in the pilot
in pilotVerifiable exampleTrustReceipt v1.0 · current corpus
JWS · Ed2551901Sample receiptReview the record and verify its signature↗02Trust Score v4.1One store's signals, with source and status↗03Purchase exampleExplore how a purchase rule responds↗
Scope: Trusteed is not a payment processor and not merchant of record: the charge is made on the merchant's own payment account, with their credentials. Nor does it replace your own records or legal obligations. See full limits
A purchase that needs approval
What happens when an order exceeds your limit?
An assistant may prepare an order above your limit or repeat a purchase by mistake. Before opening this channel, decide which actions need approval and what information you want to keep.
What if it repeats a purchase?
Identifying the assistant helps you know who is acting. You also need rules to detect repeated orders or amounts above your limit.
Agree on limits and approval points for the pilot actions.See limits and approvalsWhat changed before execution?
Price, stock, seller or terms may no longer match what was approved.
Check price and stock again in the flows that support it.See the scope of checkout controlCan you reconstruct what happened?
To review a purchase, you need to know what was requested, what was decided and which evidence is missing.
When a receipt is issued, consult the record and the missing evidence.See the sample TrustReceiptBefore, during and after the operation
Prepare the pilot and control each step.
The free audit reviews your public pages. To test purchases, Assurance identifies what is ready and what is missing. Control applies the agreed rules and Evidence lets you review the available records.
STATESavailable · in production todayin pilot · with supportplanned · no date
Beforeavailable
TRUSTEED ASSURANCESee what is ready and what is missing.
Before integrating, check which capabilities are available and which need further testing. The diagnosis helps you decide whether to start with Control or Evidence.
- Capabilities and requirements reviewed
- Signals with source and status
- Priorities for your pilot
Duringin pilot
TRUSTEED CONTROLDecide which purchases need approval.
Set limits on the assistant's actions. In the pilot, we agree on the rules your platform supports and when a request is allowed, reviewed or blocked.
- Agent context checks
- Limits, risk and human review
- Allow, review or block, with a reason
Afterin pilot
TRUSTEED EVIDENCEReview what was recorded.
On supported channels, a signed TrustReceipt lets you verify the record's integrity. Its contents depend on the evidence captured during the operation.
- Verifiable TrustReceipt
- Missing proof, made visible
- Coverage agreed for your integration
TECHNICAL DIRECTION · NOT AN AVAILABLE FEATUREThe technical substrate connects reliable data sources, decisions and proofs without replacing existing systems.
- 01Access and identityobserve
- 02Authorisation and statein pilot
- 03Policy and decisionobserve
- 04Evidence event graphplanned
Interactive decision example
A EUR 64 purchase. A EUR 420 order. What changes?
Explore three situations: an everyday purchase, an order above the limit and an unverified agent. This simulation shows fixed decisions. It does not place orders or query your store.
{ "tool": "complete_checkout", "amount": 6400, "agent_status": "verified"}policy evaluation
DECISIONALLOW WITH CONFIRMATION
The request can continue after the buyer confirms. This decision does not mean the payment has completed.
Agentic commerce
The evolution has arrived: from Online Commerce to Agentic Commerce
Rafi, our AI avatar, explains how the arrival of shopping agents is changing the rules of online commerce, and what it means for your store.
A record you can verifySee limits and verifier
Check the record of an operation.
The TrustReceipt signature lets you check who issued the record and whether it has changed. The receipt contains the available evidence for the operation, which may vary by channel and by the data captured.
The signature verifies the record's integrity. On its own, it does not prove a purchase took place or guarantee a successful dispute. The value of the evidence depends on the case and who receives it.
TRUST RECEIPTexample
receipt_id
tr_01JY7E4M8X...A92FView sample receipt data
- schema
- v1.0 · current corpus
- subject
- checkout.complete
- merchant
- verified context
- decision
- human_confirmed
- input_hash
- sha256: 4b8d...21f0
- signature
- JWS · Ed25519
In production
Individual receipts are issued in production. Coverage depends on the channel: issuance was enabled and producing at the last check (2026-10-17), and the live counters are published at /api/v1/health. This reading does not show coverage, so do not assume every purchase has a receipt. Full lifecycle tracking remains planned.
In productionv1.0Issuance and verification in production. No volume or coverage figure is provided here.
Disabledv1.1Version disabled in production
Public Trust Score v4.1
Score for merchant: demo-store (demo store, synthetic data)
65.0/ 100 · live reading from the public endpoint
The Trust Score brings together this merchant's operational reliability signals. The Agent Score from the free audit evaluates public pages. These are different measures and neither guarantees the outcome of a purchase.
CAUTIONconfidence: Limited history
There is limited history available. Interpret this score with caution; it does not guarantee future purchases.
Merchant reliabilityHistory and operationNot applicable
Agentic readinessSetup, security and policiesMeasured
Agentic evidenceOnly where a real sample existsNot applicable
Confidence with context
METHODOLOGY 4.1.0Confidence with context- The sample available, so a few cases are not overrated
- How recent the data is, because an old signal weighs less
- A confidence range, not just one number
- The declared quality of each data point
The public methodology is 4.1.0: the overall result comes from it, although the per-component breakdown may still come from an earlier engine, and each API response says which. Little history is reported as limited history, not as risk. The technical detail and API fields are on the methodology page.
How the Trust Score is measured ↗What Trusteed covers
What you can expect from Trusteed.
What we do
- We evaluate the context of the assistant's actions.
- We show the available trust signals.
- We apply limits and confirmations where they are supported.
- We issue signed records on supported channels and deployments.
What we do not do
- We are not a payment processor: the charge is made on the merchant's own account.
- We are not merchant of record.
- We do not guarantee the outcome of a dispute.
- We do not replace the merchant's legal obligations.
Questions before opening the channel
Before testing AI-assisted purchases.
Do I have to change my e-commerce platform?
You do not need to migrate your store. The pilot integrates with your existing systems and covers the actions your platform supports.
Can an authenticated agent still be a risk?
Yes. An identified assistant may repeat an order by mistake or exceed a limit. Identity is checked alongside the rules for the operation.
What does the free analysis include?
It reviews your store's public pages and returns an Agent Score with the issues found. You can start without an account. If the full report requires an email address, this is shown alongside the result. The analysis does not activate purchase controls.
Does a TrustReceipt guarantee a successful dispute?
No. It lets you verify the signed record's integrity. What it proves depends on the evidence captured, and acceptance depends on the payment provider, the process and the jurisdiction.
Your next step
Start by understanding your store.
Analyse your public pages for free. To test purchases with rules and evidence, request a pilot with your team.
For AI agents (machine-readable)
Connect your agent to a store.
MCP endpoint
POST https://trusteed.xyz/{storeSlug}/mcp/discoveryJSON-RPC 2.0: replace {storeSlug} with the slug published in merchant-index.json. Discovery needs no credential.
Discover the tools
tools/listRead the actual tool set on the selected endpoint; tools depend on each store.
Cart and checkout: /{storeSlug}/mcp/checkout. Real merchants need OAuth; demo-store uses a sandbox key and simulated payments.
If the bucket returns bucket_not_enabled_for_store, use the legacy /{storeSlug}/mcp alias. Check real-payment availability in the registry; a checkout endpoint does not prove settlement.