Skip to content
TrusteedControl and evidence for delegated operations
ESEN
Request a pilot
Trust Center

Trust is what lets merchants and agents collaborate under clear rules

Trust Center gathers verifiable status, public limits, and scope documentation. It distinguishes what's deployed from what's still in sandbox, code complete, dormant, or planned.

Request a pilot Back to the overview
Technical evidence

The v1.0 corpus lets you verify individual records.

The JWS implementation is complete and a current corpus exists. Production issuance isn't declared live until the health signal confirms activity.

Review scope and verification ↗
code completetrust-receipt+jwt
A TrustReceipt preserves signed context for a covered action. It does not by itself prove buyer intent, payment settlement or that the event occurred.
input_hash
canonical sha256
output_hash
when captured
signature
JWS · Ed25519
dormant

The hash chain and checkpoints sit behind a disabled flag; the causal graph is still planned.

Trust Score v4.1
≠a review or a guarantee

Trust Score v4.1 combines three weighted dimensions with an optional external-assurance bonus and confidence-tier caps. Each response identifies the engine used; some responses fall back to legacy v2. A legacy v2 component breakdown does not explain an overall score produced by v4.1.

Measured dataExisting, observable signalsmeasured
Not-applicable dataExplicit absence of samplenot_applicable
External assuranceOnly when applicable evidence existsbounded
RESPONSIBLE READINGThe score needs context
  • Doesn't predict commercial success
  • Doesn't certify legally
  • Doesn't replace the recipient's own decision
Explore readiness ↗
Public proof

Check the status before depending on a capability.

The canonical registry separates deployment, sandbox, code-complete, and roadmap. TrustReceipt's legal claims are still pending external approval.

STATUS
Capability statusPublic snapshot of the capability registry
Open matrix ↗
PROTO
ProtocolsSurface and limits of MCP, UCP, x402, A2A, AP2, and WebMCP
See coverage ↗
Machine-readable surface

What this site publishes so an agent can read it.

This isn't a promise: every row can be checked from the outside, without asking us for access or credentials. If a figure doesn't match what you see in the source code, the error is ours.

What's publishedStatusHow to check it
schema.org structured data51 of 70 pagesOpen the source of any page and look for application/ld+json
Questions and answers in FAQPage format20 pagesRun any page through a structured-data validator
Plain-text product summarypublishedllms.txt ↗
Canonical data in machine formatpublishedbrand-facts.json ↗
Sitemap with last-modified date171 URLssitemap.xml ↗
AI crawlers explicitly allowed11 namedrobots.txt ↗

What this doesn't mean: a machine-readable site says nothing about the maturity of what it describes. Capabilities still carry their own status, and several are in pilot or depend on integration. See the status matrix ↗

Next step

Review the proof before expanding scope

If you need to decide what evidence is missing for a business case, Assurance and Evidence help turn the review into a scoped pilot.

PilotEvaluate evidence readinessStatusReview capability maturityQuestionsAnswer common questions
Trust Center

A clearer way to open your store to AI-assisted buying

Trusteed helps stores and AI assistants understand what has been checked, what is allowed, and where human approval is still needed.

This is about making the rules clearer, not making unrealistic promises. We explain the signals so people and assistants can make better decisions.

Create free accountRead compliance information

Trust Signals

How the platform builds confidence

AgenticTrust LayerRankingEligibilityRisk controlsVerificationDomain andbusiness checksPoliciesReturns, shipping,supportOperationsCatalog, checkout,fulfillmentAgent rulesMerchant-definedguardrails

Why this is becoming real

Trust is becoming part of the infrastructure

This shift is not only about buyer behavior. Large networks and platform providers are already building the rails that make AI-assisted buying easier to trust and easier to integrate.

4.8B

Visa says its commerce network already covers 4.8 billion payment credentials, over 150 million merchant locations, and 300B+ transactions a year. Statistics sourced from Visa public documentation; citation does not imply endorsement by, or affiliation with, Visa Inc.

Visa Intelligent Commerce

20+

Google says UCP was developed with support from more than 20 ecosystem partners, including major merchants, platforms, and payment providers.

Google Developers Blog, Jan 2026

WebMCP

Chrome introduced WebMCP in early preview to let sites expose structured actions so agents can interact with more speed, precision, and reliability.

Chrome for Developers, Feb 2026

What we measure

Structured confidence, not black-box reassurance

Verification

We separate what has been checked from what still carries risk. Verification shows what is known. It does not mean every risk is gone.

Operational trust score

Trust Score v4.1 combines three weighted dimensions with an optional external-assurance bonus and confidence-tier caps. Each response identifies the engine used; some responses fall back to legacy v2. A legacy v2 component breakdown does not explain an overall score produced by v4.1.

Agent risk controls

Stores can decide when an assistant can continue, when it needs approval, and when it should stop.

1. We verify what can be verified

We check the parts that can be checked, such as store identity, business details, and public store rules.

2. We observe real operations

We look at whether store information stays current and whether the buying journey looks stable over time.

3. Merchants set the boundaries

Each store decides where assistants can move freely and where a human should still step in.

See v4.1: dimensions, bonus and legacy v2 fallback

What merchants can control

Merchants can set the tone of the relationship

We do not ask merchants to blindly trust AI shopping agents. We give them a control layer so they can define how much autonomy they want to allow and when a human should step in.

Merchant policy levers

  • Set an enforced floor for store-management agents: approval thresholds, rate limits, and credential revocation you cannot turn off
  • Define which agent behaviors require review
  • Show clear return, shipping, and support rules to agents
  • Improve visibility by improving real operational quality
  • Keep sensitive actions behind explicit confirmation when needed
See the human-approval thresholds, capability by capability

Merchant Controls

You define the rules of engagement

Merchant

Defines trust rules and review thresholds

  • Choose which agent signals matter
  • Block or review risky behavior
  • Keep autonomy where it helps sales
MCP

AI Shopping Agent

Can browse, compare, and buy within clear boundaries

  • Reads structured policies and trust signals
  • Sees when extra confirmation is required
  • Avoids merchants or flows outside policy limits

What agents can rely on

Our goal is to make trust legible to AI systems as well as to people. That means clearer semantics, machine-readable policies, and metadata that state what the platform does and does not certify.

  • Read store rules in a consistent format
  • See what has been checked and what has not
  • Understand when extra confirmation is required
  • Prefer stores with clearer and more reliable information
See TrustReceipt: signed evidence for agents

What we do not guarantee

  • Our scores are not customer review ratings
  • Our scores are not legal certification or an external audit
  • Our controls reduce risk but do not guarantee a successful transaction
  • Stores remain responsible for their legal and commercial obligations

Questions merchants and agents usually ask

Clear answers for non-technical readers and AI systems alike

Is the Store Trust Score the same as a customer rating?

No. It is an internal reliability signal based on store data and observed performance, not on public reviews.

Does verification mean a store is fully certified?

No. Verification only reflects the specific checks completed at that moment, such as identity or business review.

Can stores control how AI assistants interact with them?

Yes. Stores can define rules that review or block risky behavior before it moves forward.

Can an AI assistant rely on these signals when helping a buyer?

Yes, as decision support. The signals help with discovery, ranking, and safety, but they are not absolute guarantees.

Agent metadata

Security metadata for agents

Structured block consumable directly by AI agents to validate data and privacy constraints before acting.

{
  "storesPII": true,
  "checkoutPIIStoredTemporarily": true,
  "logsRetentionDays": 90,
  "paymentProcessing": "external",
  "dataMinimization": true,
  "merchantControlsToolExposure": true,
  "agentActionPolicy": "https://trusteed.xyz/.well-known/agent-policy.json",
  "dataProtectionRegime": "GDPR",
  "userConsentRequired": true,
  "encryptionAtRest": true,
  "encryptionInTransit": true
}

Full document at /.well-known/agent-policy.json, updated with every policy change.

Transparency by design

Scores, verification levels, and assistant controls are platform signals designed to support better decisions. They are not legal certification, an external audit, or a guarantee of commercial success.

Create free accountDeveloper docs
Trust Center | Trusteed