Agent Trust Infrastructure Platforms Compared (2026)¶
Six platforms now compete to become the trust layer for ERC-8004 AI agents. They answer fundamentally different questions - peer review quality, wallet activity, task performance, or behavioral fraud history - and those differences determine what risk each platform catches and misses.
Morgan Stanley projects nearly half of online shoppers will use AI shopping agents by 2030. The EU AI Act takes full effect in August 2026. The question of which agents can be trusted to transact autonomously is no longer a research problem.
Platform Overview¶
| Platform | Core Question | Agents Indexed | Chains | Score Range |
|---|---|---|---|---|
| ERC-8004 Native | What reviews exist? | 240K+ | ETH, BSC, Base, AVAX, Mantle | Registry only |
| RNWY | Are reviews genuine? | 185K+ | 12 chains | 0-95 (5 tiers) |
| SkyeProfile | Wallet holdings and activity? | 150K+ | Multi-registry | Dual axis |
| AXIS T-Score | Task performance quality? | Off-chain only | Off-chain only | 0-1,000 (T1-T5) |
| DJD Agent Score | Agent wallet history? | Base only | Base only | 0-100 |
| ChainAware | Owner identity and fraud history? | 240K+ | ETH, BSC, Base, AVAX | 0-1,000 (5 tiers) |
Full Capability Comparison¶
| Capability | ERC-8004 | RNWY | SkyeProfile | AXIS | DJD | ChainAware |
|---|---|---|---|---|---|---|
| Owner wallet scored | ✗ | Informational | Partial | ✗ | ✗ | ✓ Core input |
| Feeder address traced | ✗ | ✗ | ✗ | ✗ | ✗ | ✓ Unique |
| CEX feeder detection | ✗ | ✗ | ✗ | ✗ | ✗ | ✓ Positive signal |
| Rug pull history | ✗ | ✗ | ✗ | ✗ | ✗ | ✓ 1-year database |
| Honeypot token history | ✗ | ✗ | ✗ | ✗ | ✗ | ✓ Via token audits |
| Fraud prediction model | ✗ | ✗ | ✗ | ✗ | ✗ | ✓ 20M+ personas, 98% accuracy |
| Trust delegation | ✗ | ✗ | ✗ | ✗ | ✗ | ✓ Unique |
| Fleet-level farm detection | ✗ | Partial (reviewer sybil) | ✗ | ✗ | ✗ | ✓ Owner fleet database |
| On-chain readable | ✓ Registry | ✓ Base oracle | ✓ Attestations | ✗ | ✗ | Via Prediction MCP |
| Cryptographic attestations | ✗ | ✓ ES256-signed | ✓ ES256/EdDSA | ✗ | ✗ | ✗ |
| Commerce job history | ✗ | ✓ 1.7M jobs | ✗ | ✗ | ✗ | ✗ |
| Published methodology | ✓ Spec | ✓ Full weights | ✓ Provider list | ✓ 11 dimensions | ✓ | Categories only |
| Free tier | ✓ | ✓ No API key | Partial | ✗ | ✓ x402 micropay | ✓ No signup |
| MCP integration | ✗ | ✗ | ✗ | ✗ | ✗ | ✓ Prediction MCP |
Platform Profiles¶
ERC-8004 Native Reputation Registry¶
A permissionless on-chain feedback system where any wallet submits ratings for registered agents. The standard intentionally delegates scoring logic to third parties rather than defining aggregation rules - making it composable but vulnerable. One operator controlling 50 agents can manufacture high ratings between them for minimal gas cost, with no mechanism to detect it.
RNWY¶
The most established third-party platform, with the most sophisticated sybil detection in peer review systems. Uses weighted signals to identify manufactured reviews: common funder (6×), inhuman velocity (5×), sweep pattern (3×), score clustering (1×). Added owner wallet score in v1.1.0 as informational context but does not integrate it into primary scoring. Indexes 1.7 million commerce jobs across Olas and Virtuals. On-chain oracle on Base enables smart contract-level gating.
SkyeProfile¶
Multi-attestation wallet profiler aggregating nine specialized providers with cryptographically verifiable signatures - solvency, governance participation, behavioral trust, identity, security posture, compliance, performance, and settlement history. Returns independently verifiable ES256/EdDSA signatures. Delegates behavioral trust assessment to RNWY, inheriting both its strengths and its blind spot on owner-level fraud signals.
AXIS T-Score¶
Scores 11 runtime behavioral dimensions during actual task execution: completion rate, instruction adherence, data handling, transparency, error recovery, consistency, scope compliance, resource efficiency, communication clarity, security posture, and audit trail quality. Returns 0-1,000 across five tiers (T1-T5). Off-chain only. Addresses task quality rather than financial trustworthiness - orthogonal to fraud intelligence and useful as a complement, not a substitute.
DJD Agent Score¶
Narrow-focus platform scoring agent wallet addresses across seven dimensions: transaction history, partner diversity, volume patterns, account age, balance stability, activity consistency, and USDC usage. Returns 0-100 with sybil detection and gaming velocity checks. Base-only coverage. Uses x402 micropayment model (no subscriptions). Viable approach but limited scope compared to multi-chain platforms.
ChainAware¶
Inverts the approach of all other platforms by starting from the human behind the agent rather than the agent itself. Scores the owner wallet and traces the feeder address - the wallet that funded the owner. Neither RNWY's review verification nor SkyeProfile's multi-attestation model traces ownership to its funding source. A fraud operator who rotates owner wallets between campaigns is caught by ChainAware at the feeder level regardless of how clean the current owner wallet appears.
What Each Platform Catches - and Misses¶
What ERC-8004 + RNWY miss: An agent operator who has never received a review - or who deployed agents in the same block as 50 other agents - passes RNWY's sybil checks with no history to detect.
What SkyeProfile misses: The behavioral trust layer is delegated to RNWY. Owner-level fraud signals and feeder provenance are not assessed by any of the nine attestation providers.
What AXIS misses: Task performance scoring requires completed interactions. A new agent with no task history has no AXIS score. AXIS also does not address financial fraud risk - an agent that completes tasks reliably but is controlled by a rug pull operator scores well.
What DJD misses: Base-only scope excludes the majority of ERC-8004 activity. No owner wallet analysis, no feeder tracing.
What ChainAware misses: Task performance history. Commerce job volume. Cryptographic attestations. RNWY's commerce job tracking (1.7M jobs) and AXIS's runtime behavioral scoring are not replicated. For agents with established job histories, RNWY's signal is complementary rather than redundant.
The Complementary Stack¶
No single platform covers all signal types. A production-grade agent trust implementation combines:
RNWY → Sybil-resistant peer review + commerce job history
AXIS T-Score → Runtime task performance quality
ChainAware → Owner fraud history, feeder tracing, honeypot/rug pull criminal record
ChainAware is the only platform that runs before any interaction - it scores agents the moment they are registered, before they have completed any task or received any review.
Further Reading¶
- Agent Trust Infrastructure Race 2026 - full article with detailed platform deep-dives and the five differentiated signals
- Agent Trust Score - ChainAware's KYA product documentation
- AI Agent Trust & Verification - integration patterns and real-world use cases
See also: Why ChainAware | Rug Pull Detection Tools | Comparisons Overview