Snaplii's separate wallet for AI agents, explained: a ceiling is not a judgment

The receipts, dated

What the announcement confesses

The two best quotes in the announcement are the two most honest sentences written about agentic commerce this month. Xu: "You wouldn't hand a new assistant your credit card on day one. You'd give them a budget. That's what Snaplii Cash is for an AI agent, a budget, not a blank check." And: "The next question everyone asks is which card do I give it? Our answer is none of them."

"A budget, not a blank check" is the industry confessing what the ceiling is and what it isn't. A budget bounds the worst case. A budget does not judge the typical case. Inside a $50 ceiling, an agent can still buy the $45 wrong thing — the stale recommendation, the duplicated subscription, the upsell the shopper never asked for. Snaplii's wallet protects the downside; nothing in the announcement scores the purchase. That's the line the whole September authorization arc has been drawing: the six banks (Sept 22) demanded auditable records of instruction, authority, intent, and outcome; three regulators (Sept 25) said agents should not independently authorize payments; the ceiling is the instrument; scored judgment is the missing half of the instrument.

What the wallet gates — and what it doesn't

Live receipts: the gate scores the wallet

On Oct 5, 2026 ~20:25 EDT we scored two Snaplii-Cash scenarios against our live decision gate (scriptmasterlabs.com/api/harness/decide, local-heuristic-v1, bands 0.80 auto / 0.50 confirm / <0.50 escalate):

Reproduce it:

curl -sS -X POST https://scriptmasterlabs.com/api/harness/decide \
  -H 'Content-Type: application/json' \
  -d '{"state":"check","questions":[
    {"id":"q1","type":"score","scale":[0,1],
     "question":"Should the agent spend $45 of a $50 pre-funded wallet on a dinner order the shopper asked about once last month?"},
    {"id":"q2","type":"score","scale":[0,1],
     "question":"Should the agent redeem a one-time pre-funded credential for a grocery order the shopper confirmed today?"}
  ]}'

Do it yourself: add the judgment to any agent wallet

1. Keep the ceiling — a pre-funded agent balance with a hard user-set cap is the correct base layer. Snaplii proved the shape; copy it.
2. Score the instruction, not just the balance. Before money moves, score the purchase instruction on a confidence band: ≥0.80 auto-act, 0.50–0.79 confirm, <0.50 escalate. The ceiling decides how much; the gate decides whether.
3. Gate the credential issuance. One-time per-transaction credentials are the right plumbing — but issuance is where the judgment lives. If your MCP interface issues credentials automatically, put the score between the instruction and the issuance.
4. Split recommendation from transaction. If the same agent recommends and buys, score the two separately — the $120 confirmed grocery order should sail ≥0.80; the $45 stale upsell should land <0.50.
5. Log like the regulators are reading. The six-bank paper (Sept 22) wants instruction, authority, intent, and outcome records. The wallet gives you the outcome ledger; the gate's decision record completes it.

Who's building the judgment layer

We are. Decision-gated machine payments is our canonical: confidence scores gate whether an x402 machine payment fires (≥0.80 auto-pay, 0.50–0.79 confirm, <0.50 escalate). Our September 26 spending-limits builder piece is the five-part system (isolated wallet, per-payment cap, confidence gate, burn budget, kill switch) — Snaplii just shipped layers one and two as a consumer product. The gate receipts above are live on our harness endpoint today.

FAQ

What is Snaplii Cash?

A separate, pre-funded digital wallet (announced October 5, 2026) that AI agents spend from. The user loads a set amount, sets a spending limit, and the agent's purchases draw from that balance — the agent never sees the user's credit card or bank details. Cashback applies to agent-made purchases.

Which agents does it work with?

Meta's Muse today; built for other consumer agents like Instinct — one spending wallet across agents. A separate MCP-compatible Agent-to-Merchant interface issues one-time, pre-funded per-transaction credentials for developers.

How much cashback does it pay?

Per the announcement: 10% back at DoorDash, 3% at Uber Eats, up to 15% back at select brands, across 500+ brands.

Does Snaplii Cash replace a spending limit?

It is a spending limit, as a wallet: the pre-funded balance plus a user-set cap. But a cap bounds the worst case; it never scores whether any individual purchase inside the cap should fire. That's the judgment layer — and it's the one nobody has shipped yet.

Who decides whether an agent purchase should happen?

Right now: nobody scores it. Snaplii bounds it (the ceiling), isolates it (the credentials), and cheapens it (the cashback). The per-purchase confidence gate — score the instruction, band it, log the decision — is the absent layer this page builds, with live receipts.

Claim Receipts