Tiger Research has published a report examining the wallet infrastructure that autonomous AI agents need to hold, spend, and earn stablecoins without human intervention at the execution layer. The report argues that while most public attention has focused on agents executing payments autonomously, the underlying wallet stack is the actual gating constraint. The headline claim is that proper wallet infrastructure could drive a 7x revenue increase for agent-based commerce, though the report does not specify the baseline revenue figure underlying that projection. The analysis covers how agent wallets differ from standard EOA or smart-contract wallets, what architectural choices matter, and which open problems remain unsolved.
What Agent Wallets Must Do Differently
Standard crypto wallets assume a human signer who can approve transactions, manage gas tokens, and interpret UI prompts. Agent wallets break every one of those assumptions. An autonomous agent needs programmatic access to private keys, the ability to sponsor or abstract gas costs, and spend policies encoded as on-chain or off-chain rules rather than UI confirmation screens. Tiger Research frames this as a shift from interactive signing to policy-governed autonomous execution. The wallet must also handle session-based key derivation so an agent can transact for a bounded window without holding a master key indefinitely. Projects like Coinbase Agent Payments, Skyfire, and Payman are each building toward this model, but with materially different assumptions about custody boundaries, policy engines, and whether the wallet sits on-chain, off-chain, or in a hybrid configuration.
The Revenue Thesis Under Scrutiny
The 7x revenue leap claim is the kind of headline figure that demands interrogation. Tiger Research appears to be arguing that the constraint on agent commerce revenue today is not demand or protocol capability but the absence of wallet infrastructure robust enough for production deployment. That is plausible. If agents cannot reliably hold balances, enforce spend limits, reconcile settlements, and recover from failed transactions, then commercial flows cannot scale beyond demos and subsidized test traffic. The 7x figure, however, is only as credible as the baseline. Without knowing whether the starting point is a few million dollars in processed agent payments or something larger, the multiple is illustrative rather than empirical. The structural argument, that wallet infrastructure is the binding constraint, is more defensible than the specific multiplier.
Architecture Tradeoffs and Unresolved Problems
Several technical problems remain genuinely unsolved. First, key management for agents has no dominant pattern. MPC, threshold signatures, smart-contract wallets, and TEE-based key custody each offer different tradeoffs between latency, security, and recoverability. Second, agents need a way to discover service offerings and negotiate prices without hard-coded integrations, which is partly a wallet problem and partly a protocol problem. Third, spend policy enforcement must happen at a layer the agent cannot bypass, which is straightforward in a custodial model and significantly harder in a non-custodial one. Fourth, settlement reconciliation across chains, when an agent pays on one network and the service provider settles on another, introduces latency and failure modes that human users rarely encounter because they handle them manually. None of these are solved at scale.
Competitive Landscape and What to Watch
The report implicitly sketches a competitive map. Coinbase Agent Payments leverages Base and USDC, betting that a vertically integrated stack from a regulated issuer will appeal to enterprises. Skyfire is building an independent wallet and settlement layer aimed at agent-to-agent flows. Payman is taking a platform approach, abstracting the wallet entirely so developers deploy agents without managing payment plumbing. Each model has a thesis about where value accrues. What remains unclear is whether agent wallets converge on a standard, as EIP-4337 did for account abstraction, or fragment into incompatible silos tied to specific issuers and chains. The answer will determine whether Tiger Research’s revenue projection is conservative or aspirational.