DECK · 2026-07 Pre-seed / accelerator

Your agents can spend.
SpendVeto decides.

The spend-governance layer for AI agents. Payment rails move an agent's money; SpendVeto decides whether it should be allowed to move it. Open source, x402 + MCP native, working today — every claim in this deck is executed by a 267-assertion verification suite before it's published.

01 / 11

02 Problem

An agent with a wallet is a corporate card with no limit — issued to software that runs at machine speed.

AI agents now pay for tools, APIs, and data autonomously — no human, no card form. x402 alone: 169M payments across 590,000 buyers in its first year (~$50M cumulative).

Coinbase, Stripe, Visa, Mastercard, Google, and Cloudflare are all shipping agent-payment rails in 2026. Every one answers "how does the agent pay?"

Almost nobody answers the question every team asks first: "how do I stop it from overspending?"

02 / 11

03 Why now — the asymmetry

The seller side is commoditizing monthly. The buyer side is funded — but closed.

Cloudflare opened its x402 Monetization Gateway waitlist July 1, 2026 — anyone can now charge agents for anything. Stripe MPP, Nevermined, xpay, Zuplo all monetize the seller side.

Buyer-side governance is now being funded — Catena Labs raised $30M (a16z crypto, May 2026) and is filing for a bank charter — but every funded player is closed and custodial. Payman is a closed fiat platform.

The open position is empty: open-source, self-hostable, rail-neutral, MCP-native buyer-side governance. The GitLab-vs-a-bank position — SpendVeto works with any rail and any custodian, competing with none.

03 / 11

04 Product — working today, not a plan

Twelve controls between an agent and its money.

Policy enginecaps, hourly budgets, rate limits — enforced before payment
Human approvalsApprove/Deny on a live dashboard; fails closed in 30s
n-level budgetsancestor caps bind whole subtrees — IAM for money
Tool + chain scopinggrants whitelist which tools an agent may buy and which chains it may settle on
Kill switchmanual freeze + automatic runaway-burst freeze
MCP middlewaregovernance Claude cannot opt out of
Signed receiptsECDSA-signed settlements + CSV audit export
Alerts + analyticswebhooks on freeze/block; blocked-spend as a first-class metric
Enforcement proxyagents POST intents; keys never leave custody — SpendVeto on the money path
Trust scores + packsthe ledger as an agent credit file; governance as importable presets
Multichain governancechain-scoped signatures, per-chain balances, chain allowlists across 7 registered chains
Self-correcting denials + dry runsevery block returns a machine code + fix the agent can act on; dry-run any intent with zero side effects; time-boxed budgets that self-expire

Real x402 on Base Sepolia + zero-setup simulate mode with real ECDSA and replay protection. 267 end-to-end assertions: npm run verify.

04 / 11

05 Positioning

Open Policy Agent, for agent spending.

"Stripe moves money. SpendVeto decides whether the agent should be allowed to move it."

Rail-agnostic by design: x402 today; Google AP2, Stripe MPP, and Mastercard adapters on the roadmap — the policy layer shouldn't care which rail settles.

05 / 11

06 Competition

Everyone builds the roads. SpendVeto is the traffic lights.

PlayerWhat they areWhat they don't do
Coinbase x402 ecosystemThe rail itselfNo buyer-side policy, budgets, approvals
Cloudflare · Stripe MPP · NeverminedSeller-side monetizationNothing for the paying agent
Skyfire ($9.5M seed)Wallets + identity + executionNot a policy/governance SDK
PaymanClosest: limits, approvalsFiat, closed, enterprise — not x402/MCP-native
Catena LabsEnterprise agent-payment control planeClosed, institutional; no OSS, no MCP middleware
Pleo AI agents · ReimburseAIAI inside human spend/expensesGoverns humans' spend, not agents' wallets

The gap: open-source, developer-first, buyer-side governance in the stack agents already use (MCP). Skyfire's $9.5M seed — Coinbase Ventures + a16z CSX — prices this space at exactly our stage.

06 / 11

07 Market

Every agent dollar needs a spend decision before it moves.

$3–5T

TAM — global agentic-commerce spend by 2030 (McKinsey; Bain: $300–500B US). Stablecoins already move $4.5T/quarter.

69K agents

SAM — teams already paying via x402/MCP today, plus every agent-fleet team that needs governance before trusting agents with money.

$1–3M ARR

SOM, 18–24 months — OSS-led: a few hundred teams on the hosted tier (orgs, SSO, alerts, audit, enforcement proxy) + bps on governed volume.

Monetizes like usage-billing infra (Lago, Metronome, Orb): open-source core, hosted platform for money. Spend-management software (human spenders) is $26B → $56B by 2032 — agents are the next cohort.

07 / 11

08 Fit — why these funds, right now

Two RFS items in one repo.

Y Combinator — the 2026 RFS explicitly asks for agent infrastructure naming MCPs, and stablecoin infrastructure. YC now funds S26 in $500K USDC.

a16z CSX — $500K for 7%, crypto-native; already backed Skyfire one layer below us.

Alliance DAO — crypto×AI accelerator, $500K, rolling applications, decisions in 2 weeks. Next cohort starts Sept 7, 2026.

08 / 11

09 Traction plan — before applications go in

Ship in public. Walk in with usage, not promises.

Launch: GitHub OSS → Show HN, awesome-x402, MCP registries, x402 Discord. Site, deck, demo script, and listing blurbs are already built.

5 design partners running agent fleets in simulate mode — governance value lands before real money moves.

The metric nobody else can show: blocked-spend dollars — already a first-class stat — plus one testimonial from a team whose runaway agent got auto-frozen mid-loop.

09 / 11

10 Honest gaps — and what funding fixes

Prototype-grade where it should be. Real where it must be.

JSON-file storage → hosted Postgres backend is the first funded milestone.

The enforcement proxy — keyless agents, custody-side signing — ships verified as self-host today → funding builds the hosted multi-tenant version: orgs, SSO, per-team custody, SLAs.

x402 v1 packages → migrate to v2; simulate-first → mainnet Base when design partners want real settlement.

Solo founder, no production users yet → that's what the 30–60 day plan is for.

10 / 11

11 The ask

$500K to go from
verified prototype to hosted product.

Multi-rail adapters (AP2, Stripe MPP), richer anomaly models and real-time alerting, SOC 2 path, and the hosted governance dashboard that is the business.

Revanth Rajeev · [email protected]

All sources for every number: PITCH.md in the repo — McKinsey, Bain, Morgan Stanley, Forbes/BCG, Fortune BI, Coinbase disclosures.

11 / 11