New startup ideas · Fintech · Fintech
startup idea
Sigilbox
Rack appliance that hardware-signs every payment an AI agent tries to make.
Sigilbox is a co-signing appliance installed in the datacenter of any company running AI agents that spend money: each agent payment instruction must be signed by keys held in the box, which applies spend policy and an on-device behavioural model before emitting a portable attestation.
- Hardware
- Enterprise
- $10-100B market
- Network effects, winner takes most
- Global from day one
4/5
venture judge
0
similar startups, last 2 years (1 all-time)
95%
of 3 nearest real companies still alive
sector
2 public grants in this sector, none matching the idea's terms
Test it before you build it
$2,200 · 6 weeks · 20 prospects
For about $2,200 and six weeks, prove that enterprises running purchasing agents will pay for hardware co-signing before any appliance exists, and that at least one acquirer-side risk team would verify the attestation.
1Focus group: who and where
Head of AI platform, director of AI engineering, or VP of security engineering at a US company with 1,000+ employees that runs autonomous purchasing or procurement agents against a corporate card program and has eaten declines, frozen cards, or a manual review queue in the last 90 days
where to find 20 · LinkedIn Sales Navigator filtered to those titles at US companies 1,000+ employees whose recent posts or job listings mention agentic AI or AI agents; the MLOps Community Slack; the Latent Space Discord; for the verifier side, risk and underwriting leads at US payment processors reachable through Merchant Risk Council member companies on LinkedIn
2Sell first, build later
A 90-day design-partner pilot: two co-signing appliances (delivered first as the bench signer configuration) gating one agent fleet's card spend, a written spend policy implemented in the box, weekly attested-transaction reports, install starting January 2027, and an introduction of the attested traffic to the customer's own acquirer
the ask · $30,000 per pilot, credited against a year-one contract of $100,000 per two-appliance site plus $40 per agent seat per month
a real yes · A real yes is a $2,500 reservation paid through the Stripe link or a pilot agreement countersigned with a January start date and a named agent fleet; compliments, security-team enthusiasm, and unpaid proof-of-concept requests do not count
3Small experiments
The first one attacks the riskiest assumption; each ends with a number that says whether to run the next.
1. Decline-story discovery calls
$100 · 12 days
Book 20 thirty-minute calls with agent platform owners sourced from the Sales Navigator filter and the MLOps Community Slack. Script: walk me through the last time an agent-initiated card payment was declined or held, what it cost you, and what you did about it. One founder runs every call and logs incident counts and current workarounds.
keep going if · 8 of 20 describe a concrete decline or overspend incident in the last 90 days and name the internal owner of the fix
2. Verifier-side fake door
$100 · 14 days
Send a one-page attestation spec (signature format, policy fields, replay API) to 8 risk or underwriting leads at US acquirers and processors found via Merchant Risk Council member firms on LinkedIn. Ask one question: if agent transactions arrived carrying this signature, would your team consume it in sandbox to lower review friction. Log yes, no, and objections.
keep going if · 3 of 8 say they would consume the attestation in a sandbox and name the team that would integrate it
3. Priced design-partner reservation page
$150 · 14 days
Ship a one-page site: two-appliance pilot, 25 agent seats, spend policy plus attestation, install by January 2027, pilot priced at $30,000 with a $2,500 refundable reservation via Stripe Payment Link. Send it only to the 20 briefed prospects after their call, never cold. Count pilot agreement requests and paid reservations.
keep going if · 5 of 20 request the pilot agreement and 2 pay the $2,500 reservation
4. Bench co-signer demo
$1,400 · 21 days
Build a Wizard-of-Oz signer: two YubiHSM 2 units on a bench, a laptop policy engine, and a scripted agent that attempts purchases on a live virtual card, with the box approving or refusing per policy. Run it live over Zoom for the 5 hottest prospects and let them set the policy and watch a violation get blocked.
keep going if · 3 of 5 run their own policy through the bench signer and 2 proceed to a signed paid pilot
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 $2,500 refundable reservation via Stripe Payment Link, paid by the prospect's team card, followed within two weeks by a countersigned one-page pilot order form; where procurement blocks the card payment, accept a signed pilot order with a January 2027 start date and a named fleet instead set up: Stripe Payment Links ↗
what it reserves · One of four design-partner slots in the first cohort, the $30,000 pilot price locked against a $50,000 list price, and the customer's acquirer as the first attestation verifier
refund · Fully refunded on request any time before appliance install begins, one click in Stripe
target · 2 paid reservations plus 1 signed pilot order from 20 briefed prospects within 42 days
Go: build it if
8 of 20 prospects confirm a recent decline or overspend incident, 3 of 8 verifier leads would consume the attestation in sandbox, and 2 reservations of $2,500 are paid within six weeks
Kill: stop if
Fewer than 5 of 20 prospects can name a concrete incident, or zero reservations after 20 briefed prospects and 14 days of the priced page, or 0 of 8 verifiers will touch the attestation, which validates the software-only ceiling
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
One score that balances how trendy the idea is, the demand for it and its potential for 100x, with competition measured relative to every other idea in the catalog. Recent startup trends first, government priorities second.
58
Idea Score, 0-100 · raw 35.2 x 1.64
Open
competition: more crowded than 6% of ideas · headwind x0.97
+0.0
government priorities, secondary (0 matching grants)
Trend
31
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)23
- Rounds announced 2025+ in the sector25
- Sector direction (live batch)50
- 2026 trend analyst25
Demand
21
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)8
- Operator judge: real pain25
100x potential
73
Can it return a fund? The venture judge (double weight), market-size and moat axes, neighbours still alive, the technologist judge.
- Venture judge75
- Market size axis67
- Moat axis100
- Technologist judge50
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 272 ideas in the catalog; the terms matched were rack, appliance, hardware-signs, payment, tries, co-signing, installed, datacenter.
The idea in full
- What
- Sigilbox is a co-signing appliance installed in the datacenter of any company running AI agents that spend money: each agent payment instruction must be signed by keys held in the box, which applies spend policy and an on-device behavioural model before emitting a portable attestation. Processors, issuers and merchants verify that attestation for free, so an agent transaction carrying a Sigilbox signature can be accepted with lower friction than an unsigned one. It is sold as hardware per appliance plus a per-agent seat, with no money ever moving through Sigilbox.
- Why now
- Three consecutive YC batches funded companies that assume agent payment identity exists without providing it: Agentcard (S26) issuing debit cards for AI agents, Klaimee (X26) selling liability insurance for AI agents, and Mount (X26) building an AI agent insurance carrier. None of them can point to a hardware root of trust for the agent that spent the money.
- Wedge: first customer and entry point
- One enterprise already running a fleet of purchasing agents with a corporate card program that keeps getting declined: sell two appliances, then take that customer's attested traffic to their acquirer as the first verifier.
- Path to 100x
- Agent-initiated commercial spend is a slice of the $100B-scale card and B2B payments acceptance economy, and attestation is inherently two-sided: every acquirer that verifies Sigilbox makes the appliance more valuable to agent operators, and every operator makes verification worth building. One attestation format wins because verifiers will not integrate five, and the accumulated corpus of signed agent behaviour becomes the fraud model no challenger can retrain from scratch.
- Ceiling
- Agent spending stays small enough that software-only attestation is good enough and nobody buys a box.
- Closest real companies, as the generator saw them
- Agentcard issues the card and controls the credential in software; Klaimee and Mount price the risk after the fact. Sigilbox sits below all three as the physical key custody and attestation layer they each need to make their product cheaper.
- Main risk
- Cloud providers bundle agent key attestation into their existing confidential computing services and give it away.
Five judges
Each judge scores every idea in the catalog with a named rubric; the venture judge decides whether a card is shown at all (4-5 is venture-grade).
Venture investor
4/5
Attestation is genuinely two-sided and only one format wins, with Agentcard, Klaimee and Mount already proving buyers exist before the trust layer does.
Bootstrapper
2/5
Building a signing appliance for agent payments bets on a hype space where software-only attestation, as the card admits, may be good enough.
Operator
2/5
Value depends on acquirers verifying the attestation, so the appliance needs processors and issuers to move before the buyer gets anything.
Technologist
3/5
Key custody plus a behavioural model in silicon is real, but cloud confidential-computing services can bundle agent key attestation in a single release.
Risk
2/5
Value requires acquirers and issuers to verify the attestation for free, so a handful of card networks decide whether the product works.
trends
2/5
Agent-spend trend is real, but Stripe's Shared Payment Tokens and Mastercard Agent Pay already scope agent spend in software, undercutting a datacenter box.
Similar startups in the directory
Companies whose pitch matches most of the idea's terms (rack, appliance, hardware-signs, payment, tries, co-signing, installed, datacenter): 1 all-time, 0 from the last two years. Same matching as Idea Check.
BNPL infrastructure for African banks
The generator's reference companies
Real companies the model named as closest when it wrote the card, with their fate. A check mark is a company the radar could verify in its directory.
Public money in this direction
US federal grants, SBIR/STTR awards and open opportunities from the radar's public-money feed, matched to the idea's terms; the sector totals give the context.
0
grants and programs matching the idea
2
startup-relevant grants in Fintech
$809K
awarded in the sector, tracked
Market signal
What the radar sees in Fintech: new companies by cohort year, the forming YC batch, and outcomes since the February snapshot.
Fintech · 251 → 148 → 186 → 124 → 72 new companies 2022 → 2026 · 90% aliveYC S26: 11 in this cluster, 5% of the batch (was 8% in X26) (F26 is still forming: 21 listed)Since February, of 596 YC companies here: 8 acquired, 6 shut down, 45 rewrote their pitch
Design attributes
The card is one cell of a designed set: every axis below was chosen before the text was written, and the text had to realize it.
- Buyer
- Enterprise
- Business model
- Hardware
- Path to 100x
- Network effects, winner takes most
- Market size
- $10-100B market
- Capital intensity
- Capital-medium (ops, field teams)
- Speed to revenue
- Revenue in 1-3 years
- Technical depth
- Deep tech: ML, hardware, bio
- Go-to-market
- Founder-led sales
- Moat
- Data moat
- Geography
- Global from day one
- Regulation
- Unregulated
- Vibe
- Boring business
Listed under
An idea sits in its own sector and in any sector its text clearly touches.
More ideas like this
Fintech · Fintech
Bookzero
Software that closes a small company's books for a tenth of a bookkeeper's price.
Bookzero is self-serve subscription software that connects to a small business's bank feeds, invoices and payroll and produces a finished monthly close - categorized ledger, reconciliations, financial statements - for roughly $100 a month against the $800-1,500 SMBs pay outsourced bookkeepers.
Fintech · Fintech
Parbook
The self-serve secondary market where institutions trade private credit loan participations.
Parbook is a marketplace where private credit funds, insurers and regional banks list, price and settle loan participations and portfolio slices, using a standardized machine-readable data tape that Parbook generates from each seller's loan documents.
Fintech · Fintech
Perilex
An exchange where machine-priced corporate risk is listed and capital bids.
Perilex is a listing venue for idiosyncratic corporate risk that traditional carriers price badly: fleet and telematics-driven liability, compute and outage exposure, model-failure liability.
Fintech · Fintech
Countersign
Spend controls and audit trail for every dollar an enterprise's AI agents move.
Countersign sits between a company's agents and whatever rails they spend on - issued cards, wallets, stablecoin accounts - enforcing per-agent budgets, merchant allowlists and human approval thresholds before authorization, then reconciling each settled charge back to the agent run and prompt that caused it.
Fintech · Fintech
Underwatt
Underwriting software that prices AI data center risk from live power and thermal telemetry.
Underwatt sells a subscription platform to carriers, MGAs and reinsurers that prices property, business-interruption and outage risk on GPU data centers using a telemetry fleet Underwatt installs itself: power quality recorders, thermal and coolant sensors, and grid interconnect feeds at the insured sites.
Fintech · Fintech
Subrix
AI agents that settle claims between insurance carriers on one shared network.
Subrix runs autonomous agents that handle subrogation - the process where one insurer recovers claim costs from another at-fault party's insurer.
Fictional company written 2026-08-22 from MarkosWeb data; the companies, grants and numbers around it are real and tracked. Treat the idea as a research prompt, not a plan.