RNWY Trust Intelligence

RNWY is the intelligence layer for autonomous AI. It reads public on-chain data, computes trust scores, detects coordinated reviewer behavior, and issues cryptographically signed attestations. It does not gatekeep — it shows what happened and lets the relying party decide.

Philosophy: "This is not a metric. This is intelligence." RNWY rates the reviewers, not just the reviews. The evidence object — sybil clusters, funder analysis, wallet age distribution — is what separates RNWY from star-counting.

Core Endpoints

Base URL: https://rnwy.com

Authentication: Optional. Anonymous: 30 req/min. With API key: 300 req/min. Pass key as Authorization: Bearer {key} header.

GET /api/trust-check

Cryptographically signed trust attestation for any agent.

Parameters:

• id — agent_id (required)
• chain — chain slug (required)
• registry — erc8004 | olas | virtuals | sati (optional)
• threshold — minimum passing score, integer (optional)

Example:

curl "https://rnwy.com/api/trust-check?id=6888&chain=ethereum"

Response fields:

• score — overall trust score, 0–95
• tier — human-readable tier label
• pass — boolean, true if score meets threshold, false if below. No separate fail field.
• reason — plain-language explanation of the score
• badges.earned — positive signals (e.g. original_owner, long_tenure)
• badges.warnings — negative signals
• sybilSeverity — none | moderate | elevated | heavy
• sybilSignals — array of active signal names
• agentId, chain, registry
• attestation — ES256-signed envelope with expiry timestamp

Attestation: Every response includes a signed payload containing agentId, chain, registry, score, tier, badges, sybilSeverity, and timestamp. Verify signatures against the JWKS endpoint at https://rnwy.com/.well-known/jwks.json. Key ID: rnwy-trust-v2. Algorithm: ES256.

GET /api/reviewer-analysis

Wallet age distribution and sybil analysis for every reviewer of an agent.

Parameters:

• id — agent_id (required)
• chain — chain slug (required)

Example:

curl "https://rnwy.com/api/reviewer-analysis?id=6888&chain=ethereum"

Response fields:

• totalReviews — total review count
• uniqueReviewers — distinct reviewer addresses
• analyzedWallets — wallets with resolvable age data
• distribution.zeroHistory — addresses with no on-chain history
• distribution.under24h — addresses created under 24 hours before review
• distribution.under7d
• distribution.under30d
• distribution.under1yr
• distribution.over1yr
• summary.freshPct — percentage of low-history wallets (under 30 days)
• sybilFlags — array of flag description strings

Terminology: Never call low-history wallets "fresh wallets." Use "low-history," "no-history," or "spun-up" depending on context.

GET /api/compare

Side-by-side trust comparison for multiple agents, ranked by score.

Parameters:

• agents — comma-separated chain:id pairs (required)
• threshold — minimum passing score (optional)

Example:

curl "https://rnwy.com/api/compare?agents=ethereum:6888,ethereum:9380"

Response fields:

• count — number of agents compared
• threshold — threshold used
• agents[] — array sorted by score descending, each with:
• rank, score, name, chain, agentId, pass
• reviewerQuality.totalReviews
• reviewerQuality.freshPct — low-history reviewer percentage
• reviewerQuality.sybilFlags
• found — false if agent not indexed

GET /api/reviewer

Behavior profile for a specific reviewer wallet address.

Parameters:

• address — 0x wallet address (required)
• chain — chain slug (required)
• summary — true for condensed output (optional)

Example:

curl "https://rnwy.com/api/reviewer?address=0xf653068677a9a26d5911da8abd1500d043ec807e&chain=base&summary=true"

Returns velocity, sweep patterns, score clustering, funding source, and sybil signals across the entire RNWY ecosystem for that address.

POST /api/risk-terms

Counterparty risk tier and transaction signals. Returns raw data for relying parties to compute their own transaction parameters.

Body:

{ "agent_id": 16907, "chain": "base" }

Parameters:

• agent_id — integer (required)
• chain — chain slug (required)
• registry — optional

Response fields:

• risk_tier.level — 1 (lowest) to 5 (highest)
• risk_tier.label — tier label string
• recommendation — "terms" (Tiers 1–5) | "decline" (Tier 6, heavy coordinated activity only)
• signals.trust_score
• signals.address_age_days
• signals.review_count
• signals.reviewer_credibility
• signals.sybil_severity
• signals.is_original_owner — boolean
• data_coverage.signals_available / signals_total
• decline_reasons[] — array of { reason } objects, present when declining
• warning — string, present when data is partial
• auth.status — "unauthenticated" | "authenticated"

Interactive calculator: https://rnwy.com/risk-intelligence

Secondary Endpoints

# Wallet age in days
curl "https://rnwy.com/api/address-ages?address=0xf9d1...&chain=base"

# Paginated agent list with scores
curl "https://rnwy.com/api/agents?chain=base&limit=10"

# Search and filter agents
curl "https://rnwy.com/api/explorer?chain=base&limit=10"

# Registry stats
curl "https://rnwy.com/api/stats"

Trust Scoring

Formula: v2.6. Base score: 50. Cap: 95. No score reaches 100.

Scores reflect:

• On-chain registration age and tenure
• Reviewer wallet age distribution
• Sybil signal detection across reviewers
• Badge signals (positive and negative)
• Ownership history (original owner vs transferred)

Tiers by score:

risk-terms uses a separate tier vocabulary (low / moderate / elevated / high / severe / critical) for counterparty risk decisions. Do not mix the two.

Never call it a "credit score." Never call it a black box — every score shows its methodology. Adjustment, not penalty. Transparency, not judgment.

Sybil Detection

Version: v3.2 with coordination detection (v3.1).

Four signals (with score weights):

Severity levels (v2.6 nullification gate):

• heavy — review layer excluded from score entirely. Score reflects on-chain ownership signals only.
• elevated — same as heavy. Review layer excluded.
• moderate — −5 adjustment to final score.
• none — no adjustment.

When severity is elevated or heavy, the trust label renders as "Flagged — coordinated reviews detected" regardless of numeric score.

Language rules:

• Always hedged: "Indicators consistent with coordinated behavior"
• Never: "Sybil detected" or "This is fraud"
• Transparency, not judgment. Show the data; let the user decide.

Wallet Scoring

System: wallet-2.0. Dual-score model.

• Activity (signal_depth): 0–95. Depth of on-chain history.
• Risk (risk_intensity): 0–100. Concentration of risk signals.

These combine into a quadrant model. Never expose DB column names to users or in documentation. Always use the user-facing labels.

Badges

Badges are signals attached to a trust score. They are machine-readable identifiers rendered with underscores replaced by spaces for display.

Earned badges (positive): e.g. original_owner, long_tenure Warning badges (negative): surfaced from sybil and wallet signals

Replace underscores with spaces for display: original_owner → "original owner."

MCP Server

RNWY is available as an MCP tool server with 8 tools covering trust-check, reviewer-analysis, reviewer-wallet, compare, address-age, network-stats, and commerce-stats via JSON-RPC 2.0.

POST https://rnwy.com/api/mcp

SDK

npm install rnwy-sdk

import { RNWYClient } from 'rnwy-sdk'
const rnwy = new RNWYClient()
const trust = await rnwy.getTrustScore('base', 16907)

Zero dependencies. Works in Node.js 18+ and modern browsers. npm · GitHub

Machine-Readable Entry Points

Terminology Rules — Non-Negotiable