New startup ideas · Money rails for agentic commerce · Machine-to-machine and B2B payments
startup concept
Spendfence
Budgets, scoped tokens and approval policy for every agent a company runs
A control plane where finance teams issue each internal agent a budget, a merchant scope and an approval policy, then mint network-scoped tokens against those rules.
- Software subscription
- Enterprise
- Mastercard Agent Pay and Visa's Trusted Agent Protocol
6
similar startups, last 2 years (22 all-time)
yes
7 matching federal grants and programs
Direction supported by government programs and grants
Test it before you build it
$500 · 5 weeks · 20 prospects
For $500 and five weeks, learn whether controllers at mid-size software companies will pay $2,500 each for agent spend policy their existing card program does not enforce.
Riskiest assumption · A controller at a mid-size software company believes per-agent spend policy is a gap Ramp or Brex will not close for them soon, and will pay a separate vendor $2,500 today rather than wait for their card dashboard to add it.
1Focus group: who and where
Controller or VP Finance at a 100-500 person US software company where engineering already runs coding and procurement agents that buy compute, API credits and SaaS seats on the corporate card; the pain is live this month because the card statement shows agent-initiated charges nobody pre-approved.
where to find 20 · The FinOps Foundation Slack, where cloud and AI spend owners already discuss agent cost control; the Controllers Council community and its member directory; the AFP (Association for Financial Professionals) annual conference for in-person controller meetings; and warm introductions through the team's investors and former finance colleagues.
2Sell first, build later
A 60-day design-partner pilot: Spendfence turns the controller's written policy, meaning per-agent budgets, merchant scope and approval thresholds, into an enforced policy file plus dashboard in front of their existing Ramp or Brex card program, live on real agent spend within 30 days of kickoff, first cohort starting October 1, 2026.
the ask · $2,500 per 60-day pilot, then a founding price of $1,200 per month locked for year one
a real yes · A real yes is a signed one-page pilot order plus the $2,500 invoice paid by ACH before kickoff; compliments, 'we would use this', unpaid pilots and intros to procurement all count as no.
3Small experiments
The first one attacks the riskiest assumption; each ends with a number that says whether to run the next.
1. Controller gap-check calls
$100 · 10 days
Book 15 calls with controllers recruited from the FinOps Foundation Slack, Controllers Council and warm intros. Run a 5-question script: how much did agents spend on your card last month, who set each agent's limit, and what does Ramp or Brex enforce for you today. Ask them to open last month's statement on the call and count agent-initiated charges.
keep going if · 8 of 15 controllers report over $10,000 per month of agent-initiated spend and confirm no tool they use enforces per-agent budgets or merchant scope before the charge
2. Priced design-partner pitch
$300 · 10 days
Build a clickable Figma dashboard and a sample policy file (per-agent budget, merchant scope, approval threshold in YAML) populated from one prospect's real card export. Close every second call by pitching a $2,500 60-day pilot that enforces their written policy in front of their existing Ramp or Brex program.
keep going if · 5 of 12 second calls end with the controller requesting the written pilot scope
3. Signed order, paid invoice
$100 · 14 days
Send a one-page pilot order for e-signature with a fixed kickoff date and an ACH invoice for $2,500 due before kickoff. Follow up twice; anyone who stalls past 10 business days counts as a no.
keep going if · 3 controllers sign the order and pay the $2,500 invoice
4Collect a deposit up front
Tesla took $1,000 refundable reservations for the Model 3 and $100 for the Cybertruck before building either: the deposit is the measurement, not the revenue.
$2,500
per prospect, refundable
how · A design-partner pilot invoiced up front by ACH against a one-page pilot order the controller signs; a card-on-file reservation is the wrong instrument for this buyer, while an invoice against a signed order matches how mid-market finance already pays vendors and can clear in days. set up: Stripe Invoicing ↗
what it reserves · One of three slots in the first design-partner cohort, integration with their specific card program, and the $1,200 per month founding price locked for year one
refund · Refunded in full if Spendfence is not enforcing their signed policy file on live agent transactions within 30 days of kickoff.
target · 3 paid $2,500 pilot invoices from 20 controller conversations within 35 days
Go: build it if
3 paid pilots ($7,500 collected) from 20 conversations, and at least 8 of 15 discovery calls confirming over $10,000 per month of unpoliced agent spend
Kill: stop if
Fewer than 2 paid pilots after 25 conversations, or 10 of 15 controllers saying they expect Ramp or Brex to ship per-agent policy within two quarters and will wait for it
5 Scripts to run itoutreach message, landing copy, deposit terms · click to open
outreach message
You are the controller, and your engineers already let AI agents buy compute, data and SaaS seats on the corporate card. No tool you run today caps what each agent may spend, and where, before the charge happens. I am building Spendfence, a policy file and dashboard that blocks agent overruns at token issuance, in front of your existing card program. I am taking three design partners at $2,500 for a 60-day pilot. Open to a 20-minute call this week?
landing page
Give every agent a budget before it spends $2,500 buys a 60-day pilot: your policy file enforced on live agent transactions, with a $1,200 per month founding price locked for year one Book a 20-minute scoping call
deposit terms
The $2,500 pilot fee is invoiced up front by ACH and reserves one of three design-partner slots in the first cohort, with the $1,200 per month founding price locked for your first year. It is refunded in full if we are not enforcing your signed policy file on live agent transactions within 30 days of kickoff. The pilot runs 60 days from your kickoff date.
Would you run this test?
One tap. The yes-share feeds the Demand pillar of this idea's score; nobody sees who answered.
Budgets are out-of-pocket estimates for a team of one to three, US market. Size the deposit to the deal, and check the terms before taking money in a regulated line.
Scorecard
Ranked against every idea in the catalog: trend, demand and 100x potential from the corpus, competition relative to the other ideas. A generated concept has no judges or swipes yet, so its pillars use the data signals only.
58
Idea Score, 0-100 · raw 35.3 x 1.64
Active
competition: more crowded than 65% of ideas · headwind x0.67
+2.7
government priorities, secondary (7 matching grants)
Trend
49
Is the wave forming now? 2025-26 entrants vs 2023-24, rounds since 2025, the sector's live-batch direction, the 2026 trend analyst.
- Entrants 2025-26 vs 2023-24 (similar companies)62
- Rounds announced 2025+ in the sector33
- Sector direction (live batch)50
Demand
65
Does anyone want it? YC's current RFS, companies already paid for something similar, the operator judge, founders' yes-rate in decks, readers who would run the test.
- YC asks for it (current RFS: idea / sector)30
- Someone already pays (similar companies, recent / all-time)100
100x potential
36
Can it return a fund? The venture judge (double weight), market-size and moat axes, neighbours still alive, the technologist judge.
- Neighbours still alive36
Score = 100 x cbrt(Trend x Demand x 100x) x (1 - 0.5 x crowding) + government bonus (max 5), calibrated so the 95th-percentile idea scores 90 (order never changes). A geometric mean: a weak pillar cannot be papered over. Percentiles are among the 382 ideas in the catalog; the terms matched were budgets, scoped, tokens, approval, policy, control, plane, finance.
The concept in full
- What
- A control plane where finance teams issue each internal agent a budget, a merchant scope and an approval policy, then mint network-scoped tokens against those rules. Every agent purchase carries a signed policy trail the controller can audit, and overruns are blocked at token issuance rather than reviewed after the fact.
- Grounded in (2025-2026 signals)
- Mastercard Agent Pay (April 2025) and Visa's Trusted Agent Protocol (September 2025), which in 2026 became Visa Intelligent Commerce with a single integration, Intelligent Commerce Connect. Both networks joined Google's AP2 protocol in September 2025. IRBC (yc X25) pitches an 'AI-native financial institution for global businesses + agents'.
- What it rides
- Mastercard Agent Pay and Visa's Trusted Agent Protocol: Agent Pay binds an Agentic Token to one agent, one merchant scope and one consent policy, and TAP adds a verified agent identity plus an issuer-signed consent record. Spendfence is the corporate policy layer that decides what those tokens are allowed to say.
- Why now
- Both card networks shipped agent-scoped token rails between April and September 2025 and consolidated them into single integrations in 2026, but the rails enforce per-transaction scope, not company policy; nobody answers 'which of our 40 agents may spend what, where, this month'.
- Wedge: first customer and entry point
- Mid-size software companies whose engineering teams already let agents buy compute, data and SaaS seats: land as the policy file plus dashboard in front of one card program.
- Closest real companies, as the generator saw them
- IRBC (yc X25) is building the bank for businesses and agents, holding the accounts; Spendfence is bank-agnostic policy on top. Waybill (yc S26) does managed procurement for hardware teams, a human service rather than agent spend control.
- Main risk
- Card issuers bundle agent spend policy into existing corporate card dashboards before a standalone control plane earns its seat.
Similar startups in the directory
Companies whose pitch matches most of the concept's terms (budgets, scoped, tokens, approval, policy, control, plane, finance).
The Agent Control Plane for IT teams
The control plane for AI agents.
Guest-travel & expense management with intelligent policy enforcement and seamless user experience
The spend control layer for AI agents
The default way of running on-device AI at Scale
Secure access control and authorization platform for agent workflows
Self-driving observability, evals, and gateway for AI agents
Workflow Automation for Construction Finance
The Finance Control Centre for African Businesses
Adaptive is a cybersecurity platform that is reimagining access for an AI native world
Dynamic access controls for all data and databases
Joint banking for millennial couples.
Public money in this direction
US federal grants and open opportunities matched to the concept's terms.
National Science Foundation · SBIR Phase I · $305K
NIH / NIDDK · SBIR phase I · $226K
National Science Foundation · Secure &Trustworthy Cyberspace · $150K
National Science Foundation · Secure &Trustworthy Cyberspace · $150K
National Science Foundation · Secure &Trustworthy Cyberspace · $384K
National Science Foundation · Secure &Trustworthy Cyberspace · $365K
National Science Foundation · Networking Technology and Syst · $675K
Other concepts in this collection
- QuadrailOne endpoint that speaks ACP, UCP, AP2 and both card network agent protocols
- CartproofContinuous conformance testing for merchant agent checkout endpoints
- TokentraceReconciliation built around scoped payment tokens for PSPs handling agent orders
- CountersignDispute defense for agent purchases using issuer-signed consent records
- DutylineLanded cost inside the agent checkout response for cross-border orders
- ConsentryVerification gateway that tells merchants which agent orders to trust
- KeyfleetCredential lifecycle for every AI agent a company lets spend money
- MintgatePolicy engine deciding when an agent gets a scoped payment token
- TrailproofAn audit agent that reconstructs consent for every dollar your agents spent
- HoldpointHuman approvals that become signed consent records for agent purchases
- MeterlinePer-agent stablecoin allowances and kill switches for machine-to-machine payments
- VetgateOne decision point for banks verifying agent credentials across every rail
- AttestaChargeback evidence packets built from agent consent records
- ReversioRefund rails for purchases made with expired scoped tokens
- TallybridgeReconciles agent-originated orders against ERPs and the general ledger
- VerdiktIssuer-side triage for disputes on agent-initiated transactions
- RefluentRefund and reconciliation layer for stablecoin agent payments
- DutybackRecovers duties and tax when cross-border agent orders come back
- NetformNetting and clearing for millions of tiny machine-to-machine payments
- CovenorYour AP agent negotiates terms and settles invoices with their AR agent
- FloatworksTreasury operations for the stablecoin float that funds agent budgets
- Recourse LabsDispute and refund rails for payments agents make on their own
- CharterlineLicensing and exam readiness software for stablecoin issuers under the GENIUS Act
- RelayproofTravel rule data for stablecoin transfers that an agent initiated
- SievewireSanctions and AML screening built for machine-speed agent payments
- AttestlyOne evidence vault for agent checkout consent records and scoped tokens
- AmletAML program operations run by agents for seed-stage stablecoin PSPs
- YieldgateProof that stablecoin rewards are activity linked, not prohibited interest
- ClearquoteLanded cost inside the agent's checkout quote, duties and taxes priced before the token
- DrawpointAn agent that files duty refunds when cross-border agent orders come back
- PegwiseStablecoin settlement between the agent's currency and the merchant's, with locked FX
- DutygraphThe customs knowledge base agents query before buying across a border
- PortruleSpend policy for purchasing agents that buy from foreign suppliers
- AttestoRegulator-ready evidence for every cross-border stablecoin settlement an agent triggers
- ParflowRedemption operations API for stablecoin issuer programs
- StablebooksTreasury policy and reconciliation layer for companies holding stablecoins
- CorridorlinePayout orchestration across stablecoin rails with compliance evidence built in
- RegimarkAgent service that builds dual GENIUS and MiCA compliance evidence for issuers
- RewardrailCompliant activity-linked rewards engine for stablecoin issuers and platforms
- FeedstoneAn agent that builds and maintains your UCP and ACP catalog endpoints
- TracelightAttribution and revenue reporting for orders placed by AI agents
- GatewickThe traffic gate that tells buying agents from scraping bots
- ClausemarkPublish return, shipping and warranty policies agents can read and bind to
- CounterproofChargeback evidence built from agent consent records and scoped tokens
- StockproofReal-time inventory and price truth for agent catalog queries
Fictional concept generated 2026-08-26 by claude-fable-5 from the collection's brief and MarkosWeb data. Treat it as a research prompt, not a plan.