Skip to content
Trust infrastructure for autonomous agents

Know which
agents to trust.

Turn verifiable behavior into explainable trust scores and risk signals your systems can use.

Check an agent Read the docs
Open sourceExplainableDeterministicEvidence-led
Agent trust profile
Illustrative response · demo data
GET /score
Demo profile ready
Agent identity
GD64YIY3…BHU5A
78/ 100
Low risk
confidence 0.84
Why this score?5 inputs
Longevity82
Established history
Activity74
Consistent use
Velocity86
No burst pattern
Network69
Limited delegation
riskLevel: lowevidence: 2 peers · 6 edgeschecksum: sha256…Inspectable output
Verifiable behaviorExplainable scoresDeterministic decisionsOpen source
From unknown to understood

Identity is a starting point.
Evidence makes it useful.

Autonomous systems need more than an address before they delegate work, pay, or extend a limit.

01 · unknown agent
GD64YIY3…BHU5A
An identifier without context
02 · observable evidence
HistoryRelationshipsReputation
Signals your application can inspect
03 · explainable decision
Trust 78 Low risk
A verdict with reasons attached
Three ways to use it

Trust is a product primitive.

Use the same evidence layer to rank, verify, and approve autonomous interactions.

01Score

See the signal behind the number.

Turn activity, age, volume, velocity, and balance into a 0–100 score with reasons attached.

78
trustScore / 100
02Verify

Understand the network around an agent.

Inspect delegation, reputation, relationships, and Sybil-like indicators before you act.

relationship → depth → evidence
03Underwrite

Make a decision your system can defend.

Combine trust signals into deterministic risk decisions, capacity estimates, and limits.

riskLevel
low
→
action
approve
How it works

One identity.
Four clear steps.

No opaque classification. The output carries the evidence your system needs to act.

01

Provide identity

A wallet or agent identifier.

02

Analyze signals

History, relationships, and behavior.

03

Evaluate risk

Deterministic factors and thresholds.

04

Return a verdict

JSON your application can inspect.

For developers

From agent ID to trust verdict in one request.

Use HTTP, TypeScript, or Python. Start locally, inspect the response, and keep the decision in your own system.

Read the docs 
const result = await client.getScore(wallet)

// { trustScore: 78, riskLevel: "low", ... }
→
structured response
78low risk
explanationattached
checksumsha256…
Trust the method

Evidence before confidence.

Open source, deterministic scoring, and verifiable on-chain inputs make the boundary visible.

Algorand provides the observable activity and relationship history; Agent Passport turns it into auditable application signals.

Read security notes
01

Observable inputs

Activity and relationships are derived from the configured Algorand sources.

02

Inspectable output

Scores include component values, reasons, risk levels, and source context.

03

Explicit limits

Signals do not prove a human identity or guarantee future behavior.

04

Open by default

Run the service yourself and inspect the algorithms, contracts, and tests.

Agent Passport

Know who you’re dealing with.

Evaluate an agent before your system trusts it.

Check an agent Read the docs