LAST UPDATED: MARCH 22, 2026
How RNWY translates on-chain trust signals into risk-informed transaction parameters. Every formula published. Every calculation reproducible.
Counterparty risk intelligence for marketplaces, escrow providers, and agent orchestrators.
RNWY provides counterparty risk intelligence for AI agents registered on ERC-8004 and other on-chain registries. Marketplace operators, escrow providers, and agent orchestrators use this data to inform their own transaction decisions.
This is informational risk intelligence. RNWY does not hold funds, enforce terms, or make binding decisions. All outputs are derived from public on-chain blockchain data. Off-chain behaviors, private agreements, and real-world identity are not captured. The relying party is solely responsible for all transaction decisions made using this data.
The principle is Transparency, Not Judgment. Every signal, every weight, every formula is published on this page. A third party can independently reproduce any output from the API. We show the math. You decide.
POST /api/risk-terms accepts an agent ID and chain, and returns the agent's risk tier, underlying trust signals, and a reference to this methodology page. The response includes every raw signal so relying parties can independently compute their own transaction parameters using the formulas published below.
You send: agent_id (required), chain (required, e.g., "base"), registry (optional, defaults to "erc8004").
You receive: risk tier (level, label, description), all six input signals with raw values, methodology version and URL, and a legal disclaimer.
Full endpoint documentation, authentication, and rate limits are on the API page.
Sample Response
{
"recommendation": "terms",
"agent_id": 16907,
"agent_name": "Wolfpack",
"chain": "base",
"chain_id": 8453,
"registry": "erc8004",
"explorer_url": "https://rnwy.com/explorer/base/16907",
"risk_tier": {
"level": 3,
"label": "elevated",
"description": "Elevated counterparty risk; limited trust signals established",
"computed": true
},
"signals": {
"trust_score": 54,
"score_available": true,
"sybil_severity": "moderate",
"address_age_days": 142,
"is_original_owner": true,
"review_count": 12,
"reviewer_credibility": "medium"
},
"data_coverage": {
"address_age": true,
"sybil_analysis": true,
"reviewer_credibility": true,
"signals_available": 6,
"signals_total": 6
},
"warning": null,
"methodology": {
"version": "1.0.0",
"url": "https://rnwy.com/risk-intelligence",
"description": "Tier definitions, modifier weights, and all formulas are published at the methodology URL."
},
"disclaimer": {
"status": "INFORMATIONAL_ONLY",
"text": "..."
}
}When signal coverage is incomplete, the response includes a warning string and the data_coverage field shows which signals are unavailable. Some chains do not yet have full data provider support for address age or sybil analysis. The trust score in these cases is computed from available signals only; the warning makes this explicit.
Every agent is assigned one of six risk tiers based on its trust score. Tiers 1 through 5 are score-derived. Tier 6 is triggered by absolute risk indicators and returns a decline recommendation with reasons.
| Tier | Label | Trust Score | Description |
|---|---|---|---|
| 1 | low | 75–95 | Low counterparty risk; strong on-chain trust signals |
| 2 | moderate | 60–74 | Moderate counterparty risk; solid but not extensive trust history |
| 3 | elevated | 45–59 | Elevated counterparty risk; limited trust signals established |
| 4 | high | 25–44 | High counterparty risk; few trust signals, elevated caution recommended |
| 5 | severe | 0–24 | Severe counterparty risk; insufficient trust history for confidence |
| 6 | Critical | n/a | Triggered when indicators consistent with severe coordinated activity are identified. Returns decline recommendation. |
Tier 6 is not score-derived. It fires when absolute risk triggers are detected. Currently, the only trigger is detection of indicators consistent with heavy coordinated activity. The response includes the specific reasons for the decline recommendation. No parameters are computed for declined agents.
The API returns six raw signals for every agent. These are the inputs to the risk tier and parameter calculations.
A composite score computed nightly from on-chain data. Base of 50. Incorporates on-chain activity, feedback quality, address history, and ownership continuity. This is the primary signal that determines tier placement. When the score is unavailable, the API returns score_available: false and assigns Tier 5 as a conservative default.
One of five levels: none, low, moderate, elevated, or heavy. Derived from RNWY's pattern detection pipeline, which analyzes reviewer wallet clustering, funding patterns, and temporal coordination. Levels reflect the strength of detected indicators, not confirmed verdicts. When indicators consistent with heavy coordinated activity are identified, the API returns a decline recommendation. A null value means pattern analysis has not yet run for this agent.
The age of the wallet address that owns the agent, computed from the address's first on-chain transaction via Alchemy. Addresses under 30 days old are flagged as low-history wallets. Addresses over 365 days old receive a mature address signal. A null value means the address age could not be determined. The API distinguishes between confirmed low-history and data unavailable.
Whether the current wallet is the original registrant or received the agent via transfer. ERC-8004 agents are ERC-721 NFTs; they can change hands. An agent that has been transferred carries a different risk profile. When ownership data is unavailable, the API conservatively reports the agent as transferred.
The total number of on-chain feedback entries for this agent. Agents with fewer than 3 reviews have an insufficient feedback signal; the data exists but there isn't enough of it to draw conclusions.
Derived from the percentage of reviewers using low-history wallets. High means 80%+ of reviewers have established wallet histories. Medium means 40–80% established. Low means under 40%. Null means insufficient data (fewer than 5 reviews).
Relying parties can use the signals above with the formulas below to compute recommended transaction parameters. These are the exact formulas used by the interactive calculator at the bottom of this page.
| Tier | Collateral | Max TX | Escrow | Evaluator |
|---|---|---|---|---|
| 1. low | 15% | $500,000 | 24h | optional |
| 2. moderate | 35% | $50,000 | 48h | optional |
| 3. elevated | 55% | $10,000 | 72h | recommended |
| 4. high | 75% | $2,000 | 96h | required |
| 5. severe | 100% | $500 | 120h | required |
All modifier deltas are summed, then applied once: modified_collateral = base_collateral × (1 + sum_of_deltas)
| Signal | Condition | Collateral Delta | Max TX |
|---|---|---|---|
| Indicators: elevated | indicators = elevated | +25% | Halved |
| Indicators: moderate | indicators = moderate | +10% | — |
| Indicators: low | indicators = low | -10% | — |
| Indicators: none | indicators = none | -5% | — |
| Address age < 30d | Low-history wallet | +20% | Halved |
| Address age > 365d | Mature address | -10% | — |
| Transferred | Not original owner | +15% | — |
| Reviews < 3 | Insufficient feedback | +10% | — |
When multiple signals halve the max transaction value, the halving compounds. Each halving signal multiplies the divisor by 2. One signal = ÷2. Two signals = ÷4. The divisor and every contributing signal are shown in the API response and the calculator breakdown.
When a transaction value is provided, collateral is adjusted using logarithmic scaling:
scaling_factor = max(0.5, 1 + 0.1 × ln(tx_value / 1000)) effective_collateral = modified_collateral × scaling_factor
The reference value is $1,000 (scaling factor = 1.0, no adjustment). Below $1,000, collateral decreases slightly. Above $1,000, it increases at a decreasing rate. The scaling factor is floored at 0.5; no transaction gets more than a 50% collateral reduction from value scaling.
Agents within 3 points below a tier boundary receive blended terms to prevent gaming at boundaries. Collateral and escrow timeout are interpolated between adjacent tiers. Max transaction value and evaluator recommendation use the stricter (lower) tier.
blend_weight = (boundary - score) / 3 blended = higher_tier × (1 - blend_weight) + lower_tier × blend_weight
At the boundary itself (weight = 0), the agent gets full higher-tier terms. At 3 points below (weight = 1), the agent is fully in the lower tier with no blend.
If the base evaluator recommendation is "required" AND reviewer credibility is "high" AND the trust score is in the top 50% of the tier range, the evaluator recommendation downgrades one step to "recommended." This is the only evaluator adjustment. The rule is deterministic; the same inputs always produce the same output.
Minimum collateral: 10%. No combination of favorable signals reduces collateral below this floor.
Maximum collateral: 150%. If modifiers and scaling push collateral above this ceiling, the response shows the calculated value and the capped value separately, with a warning suggesting the relying party evaluate whether the transaction should proceed.
Current version: 1.0.0. Every API response includes the methodology version. Changes follow semver:
Major version (e.g., 2.0.0): tier boundary shifts, new decline triggers, structural changes to the response shape.
Minor version (e.g., 1.1.0): modifier weight adjustments, scaling parameter tuning, new optional response fields.
Patch version (e.g., 1.0.1): documentation corrections, no behavioral changes.
Relying parties can pin to a specific methodology version in the API request. Previous versions remain documented in the changelog below.
v1.0.0 (March 22, 2026); initial release. Five score-derived tiers plus critical decline tier. Six input signals. Additive modifier stacking. Logarithmic value scaling. Buffer zones. Evaluator softening. Collateral floor 10%, ceiling 150%.
This calculator uses the exact same formulas described above, running client-side in your browser. Adjust the inputs and see how each signal affects the recommended parameters. The step-by-step breakdown shows every modifier and calculation.
Methodology v1.0.0. All computation runs client-side using the published formulas.
RISK TIER
3. elevated
RECOMMENDED COLLATERAL
52.3%
MAX TRANSACTION
$10,000
ESCROW TIMEOUT
72h
EVALUATOR
recommended
Step-by-Step Breakdown
All outputs from the RNWY Transaction Risk Intelligence API and this methodology page constitute risk intelligence for informational purposes only. These are not binding recommendations, financial advice, or credit decisions. The relying party is solely responsible for all transaction decisions made using this data.
Outputs are derived exclusively from public on-chain blockchain data. Off-chain behaviors, private agreements, and real-world identity are not captured. Data may be temporarily unavailable, delayed, or contain errors. RNWY does not guarantee the accuracy, completeness, or timeliness of any output. The relying party is solely responsible for all transaction decisions made using this data.
Terms of Service →