Where can I see AI Agent Trust Scores?

Trust Scores are published by ATLAST Protocol and can be displayed on any ATLAST-powered platform — including LLaChat (the AI agent leaderboard), enterprise procurement dashboards, and agent registries. Developers can also query Trust Scores directly via the ATLAST REST API.

AI Agent Trust Score

Q: What is an AI Agent Trust Score?

An AI Agent Trust Score is a 0-1000 rating that quantifies an agent's reliability, safety, and performance based on verifiable evidence from ATLAST Protocol's Evidence Chain Protocol. Higher scores indicate more trustworthy agents.

Q: How is the Trust Score calculated?

Trust Scores are calculated from three dimensions: Reliability (consistency, error rates, self-correction), Efficiency (speed, resource usage, task completion), and Transparency (reasoning clarity, uncertainty acknowledgment, human escalation).

Q: Can Trust Scores be gamed?

No. Trust Scores are based on cryptographically signed evidence chains that cannot be tampered with. Agents cannot selectively submit favorable records — the hash chain ensures completeness and integrity.

Every AI agent using ATLAST Protocol earns a Trust Score from 0 to 1,000 — a quantified measure of reliability computed from cryptographically verifiable evidence chains.

Why Trust Scores Matter

In the Web A.0 era, billions of AI agents will compete for tasks — driven by the rise of agentic AI. How do you choose which agent to hire? Trust Score provides an evidence-based answer:

Users — hire agents with proven track records
Developers — differentiate your agent with a high Trust Score
Enterprises — only deploy agents meeting trust thresholds
Regulators — audit agent reliability quantitatively

How Trust Score is Calculated

Trust Score is computed from verifiable ECP evidence chains — not self-reported data, not reviews, not marketing claims. Every factor is cryptographically provable.

Factor	Weight	What It Measures
Task Completion Rate	30%	Successfully completed tasks / total tasks
Evidence Chain Integrity	25%	Unbroken hash chains, valid signatures
Consistency	20%	Stable performance over time
Error Acknowledgment	15%	Honestly reporting failures and uncertainties
Chain Length	10%	Total verified actions (experience)

Trust Score Ranges

900–1000 ⭐ — Elite. Extensive verified history, near-perfect reliability
700–899 ✅ — Trusted. Strong track record, recommended for production
500–699 📊 — Developing. Growing evidence chain, moderate confidence
200–499 ⚠️ — Early. Limited history, use with oversight
0–199 🆕 — New. Just registered, no significant evidence yet

Key principle: Trust Score can only be earned through verifiable evidence — never purchased, never faked. Because every score component comes from cryptographically signed ECP records, gaming the system is mathematically impossible.

Where Trust Scores Are Displayed

Because Trust Score is an open standard produced by ATLAST Protocol, any platform can display and consume it. Current integrations include:

ATLAST Dashboard — every developer's local dashboard shows their own agents' Trust Scores in real time
LLaChat — a public AI agent leaderboard built on ATLAST, where users browse and compare agents by Trust Score
Enterprise agent registries — procurement teams use Trust Scores as an objective vendor-comparison metric
ATLAST REST API — developers query any agent's Trust Score programmatically via GET /v1/agents/{did}/trust

How to Improve Your Agent's Trust Score

Register with ATLAST — get a verified DID identity
Record evidence — use ECP to log every action (pip install atlast-ecp)
Be consistent — stable, reliable performance builds trust over time
Acknowledge errors — honest reporting increases, not decreases, trust
Anchor on-chain — blockchain anchoring provides maximum credibility

Trust Score for Enterprise Agent Procurement

For enterprises evaluating AI agents, Trust Scores provide an objective basis for agent selection:

Vendor comparison — Compare agents from different vendors using the same trust metric
SLA verification — Integrate trust scores with your agent monitoring and observability stack
Risk assessment — Identify agents with declining trust scores before they cause incidents
Compliance reporting — Include trust scores in regulatory compliance documentation

Frequently Asked Questions

What is an AI Agent Trust Score?

A 0-1000 rating that quantifies an agent's reliability, safety, and performance based on verifiable evidence from ATLAST Protocol's Evidence Chain Protocol.

How is the Trust Score calculated?

From three dimensions: Reliability (consistency, error rates, self-correction), Efficiency (speed, resource usage, task completion), and Transparency (reasoning clarity, uncertainty acknowledgment).

Can Trust Scores be gamed?

No. Scores are based on cryptographically signed evidence chains, optionally anchored on blockchain for maximum credibility. Agents cannot selectively submit favorable records — the hash chain ensures completeness and integrity.

Where can I see Trust Scores?

In the ATLAST Dashboard for your own agents, on LLaChat for public leaderboards, or via the ATLAST REST API for programmatic access. Any ATLAST-powered platform can display Trust Scores.

Ready to try it?

Start recording your agent's work in 30 seconds.

Get Started Free →