MarkosWebWeb intelligence for the agentic economy

New startup ideas · Money rails for agentic commerce · Cross-border commerce inside agent checkout

startup concept

Dutygraph

The customs knowledge base agents query before buying across a border

A machine-readable API of HS classifications, duty rates, de minimis thresholds, restricted-party rules and import permissions, versioned daily and served in the shape agent frameworks expect.

11

similar startups, last 2 years (48 all-time)

yes

8 matching federal grants and programs

Direction supported by government programs and grants

Test it before you build it

$500 · 4 weeks · 20 prospects

For $500 and four weeks, prove agent builders will pay $250 deposits and $199 a month for a customs API their buying agents query before checkout.

Riskiest assumption · Teams shipping cross-border buying agents treat import legality and landed cost as a blocking problem they will pay for now, rather than skipping foreign listings or getting by with existing landed-cost tools built for human checkouts.

1Focus group: who and where

A founder or founding engineer at a 2-10 person, seed or YC-stage team shipping a purchasing or shopping agent that can buy from EU or UK merchants, who has already had an order stopped at customs or quietly makes the agent skip foreign listings.

where to find 20 · The YC Startup Directory searched for shopping, purchasing and commerce agents; AI Tinkerers meetups in San Francisco and New York; the LangChain Discord and the Model Context Protocol GitHub discussions, where agent-checkout tooling is actually being built.

2Sell first, build later

A first-cohort API key for the US-imports-from-EU-and-UK corridor: HS classification, duty rate, de minimis, restricted-party and admissibility answers with a signed daily change log, keys shipped 30 days after deposit, with answers human-served within 24 hours during the pilot at the same price on the same contract.

the ask · $199 per month founding tier, locked for 12 months, after a $250 refundable deposit.

a real yes · A paid $250 deposit is a yes. GitHub stars, 'we would totally use this', and requests for a free trial are not.

3Small experiments

The first one attacks the riskiest assumption; each ends with a number that says whether to run the next.

  1. 1. Blocking-problem interviews

    $100 · 10 days

    Book 15 calls with builders of cross-border purchasing agents found through the YC directory, AI Tinkerers and the LangChain Discord. Ask what their agent does today when a listing is foreign - check, guess or skip - and what one stranded shipment cost them in refunds and support time.

    keep going if · 9 of 15 say import legality or landed cost blocks checkout or causes failed deliveries, and none has a working automated answer

  2. 2. Priced spec pre-order

    $150 · 12 days

    Publish a docs page with a complete OpenAPI spec for the US-from-EU/UK endpoint - example requests, the signed change-log format, the $199 a month founding price - and a $250 deposit checkout for one of 10 first-cohort keys; a self-serve payment link fits here because developer buyers expect to pay online without a sales call. Send it to all 15 interviewees plus 10 more builders from the same channels.

    keep going if · 5 paid $250 deposits from 25 offers

  3. 3. Human-backed accuracy pilot

    $250 · 12 days

    Give the first 3 depositors a pilot endpoint the founders answer by hand within 24 hours from the USITC HTS, EU TARIC and UK Trade Tariff sources - hand-serving real queries is deliberate, because the card's stated risk is whether the data can be kept right, and this measures real query volume and error rate at once. Pay a licensed customs broker to spot-check 25 answers.

    keep going if · 150+ real queries across 3 teams in 12 days, with at most 1 of 25 spot-checked answers disputed

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.

$250

per prospect, refundable

how · A $250 refundable reservation taken through self-serve checkout on the docs page, because developer buyers evaluate a spec and pay online without a sales cycle; the deposit converts to a credit on the first invoice at launch. set up: Stripe Invoicing ↗

what it reserves · One of 10 first-cohort API keys for the US-from-EU/UK corridor, the $199 a month founding price locked for 12 months, and a vote on the first 5 endpoint additions.

refund · Fully refunded on request any time before the key ships, and automatically if the key misses the 30-day ship date.

target · 5 paid deposits from 25 offers within 25 days

Go: build it if

5 paid deposits from 25 asks, 150+ real queries served in the pilot, and 24 of 25 broker-spot-checked answers confirmed correct.

Kill: stop if

Fewer than 3 deposits from 25 asks, 9+ of 15 interviewees saying they simply skip cross-border listings and lose nothing by it, or 3+ of 25 spot-checked answers wrong.

5 Scripts to run itoutreach message, landing copy, deposit terms · click to open

outreach message

Saw you are shipping a purchasing agent that can buy across borders. When it finds a product in the EU or UK, does it know before checkout whether the item can legally enter the US and what it lands at? I am building Dutygraph: a versioned customs API - HS codes, duty rates, de minimis, restricted-party rules - served the way agent frameworks expect, with a signed change log your agent can cite. Ten first-cohort keys at $199 a month locked for a year. Got 20 minutes this week?

landing page

The customs answer your buying agent needs before checkout Founding tier $199 a month locked for 12 months; a $250 refundable deposit reserves one of 10 first-cohort API keys Reserve your API key - keys ship in 30 days

deposit terms

Your $250 deposit reserves one of 10 first-cohort API keys for the US-from-EU/UK corridor, shipping within 30 days, and locks the founding price of $199 a month for 12 months. It is fully refundable on request any time before your key ships, and refunded automatically if we miss the ship date. At launch the deposit is credited to your first invoice.

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.

42

Idea Score, 0-100 · raw 25.9 x 1.64

Crowded

competition: more crowded than 80% of ideas · headwind x0.60

+3.8

government priorities, secondary (30 matching grants)

Trend

52

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)74
  • 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

15

Can it return a fund? The venture judge (double weight), market-size and moat axes, neighbours still alive, the technologist judge.

  • Neighbours still alive15

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 customs, knowledge, base, query, buying, machine-readable, classifications, duty.

The concept in full

What
A machine-readable API of HS classifications, duty rates, de minimis thresholds, restricted-party rules and import permissions, versioned daily and served in the shape agent frameworks expect. Agent builders and commerce platforms call it at decision time so a buying agent knows whether an item can legally enter the destination country and at what cost, before checkout starts.
Grounded in (2025-2026 signals)
UCP 'by March 2026 it covered carts, live catalog queries and loyalty identity'; YC RFS Fall 2026 'Self-Maintaining APIs'; keyword coverage shows 'customs' at 7 companies in the last 2 years against 31 for 'fraud', so customs data as a product is thinly covered; Trava (yc W25, 2025) and Tarifflo Inc. (yc S26, 2026) show accelerators funding trade-compliance entries in both years.
What it rides
UCP: one checkout protocol for agents. UCP's live catalog queries give agents product data but no importability data; Dutygraph is the missing lookup an agent makes between finding the product and requesting the token.
Why now
Customs rules changed constantly through 2025-2026 tariff volatility while UCP reached live catalog queries in March 2026, so agents can now find foreign products faster than they can learn whether importing them is legal; the current YC RFS asks for Self-Maintaining APIs, which is exactly what a daily-versioned customs API is.
Wedge: first customer and entry point
Three agent-builder teams shipping cross-border purchasing agents; sell a single-corridor endpoint, US imports from the EU and UK, with usage pricing and a signed change-log so the agent can cite the rule version it relied on.
Closest real companies, as the generator saw them
Trava builds AI agents for global trade compliance and Tarifflo Inc. does supply chain compliance, both selling workflows to shippers and compliance teams; Dutygraph sells the underlying data as an API to the machines doing the buying. Donkey embeds this knowledge inside its own trading operation rather than exposing it.
Main risk
Keeping tariff and restriction data correct across jurisdictions is a grind, and one wrong answer that strands a shipment destroys trust in the whole dataset.

Similar startups in the directory

Companies whose pitch matches most of the concept's terms (customs, knowledge, base, query, buying, machine-readable, classifications, duty).

  • Proxyc F25 · 2025 · Vertical AI agentsalive

    AI technical support for complex physical products

  • Quantstructyc W25 · 2025 · Developer toolsalive

    AI documentation engineer – test & autoimprove stale product docs

  • Intropicef EF London 2019 · 2019alive

    Providing data refineries for investment management firms.

  • Mindoriyc S16 · 2016 · Commerce and marketplacessite down

    Mindori builds white-label voice search for e-commerce apps.

  • Bloomsbury AIef EF London 2016 · 2016acquired

    AI for automating customer support workflows.

  • Sentient OSyc F26 · 2026 · Horizontal AI assistantsalive

    On-device AI that knows your entire life and does your work overnight.

  • Donkeyyc S26 · 2026 · Commerce and marketplacesalive

    The AI-native trading company: China factory-direct for US importers

  • Vendoyc S26 · 2026 · Developer toolsalive

    Let your users build their own features on top of your product

  • Woodstockplugandplay PnP 2026 · 2026 · Fintechalive

    1000円からはじめられる米国株ソーシャル投資アプリ「woodstock.club」の提供

  • Heavispeedrun SR006 · 2026 · Vertical AI agentsalive

    AI workforce for heavy vehicle mechanics. Two 2x founders who raised $200M and $75M prev

  • Alpha Researchyc X25 · 2025 · Agent infrastructurealive

    Open source, agentic knowledge bases for all of humanity's knowledge.

  • Promptlessyc W25 · 2025 · Developer toolsalive

    An AI teammate that automatically updates your customer-facing docs

Public money in this direction

US federal grants and open opportunities matched to the concept's terms.

Other concepts in this collection

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.