New ideaAgent Services

Agent Door

A one-snippet 'door' that lets a small merchant or site decide, per personal AI agent (Muse, Dots, Grok Bot, Instinct), whether to allow, block or charge it, and serves agents a clean machine-readable menu of bookable actions (availability, catalog, quotes).

3 weeks ago · Scored 20 Oct 2026 · ID HBV-261020-H07

Summary

BriefingHypothesis

BET / HYPOTHESIS. The Verge (Oct 17 2026): Yelp only allows agents that pay for data licensing, Delta blocks agent bookings, JPMorgan and Schwab block agents, while Meta partnered with Walmart, Gap and Sephora and Shopify opened its catalog to more agents. Big sites are choosing; small sites have no easy tool.…

Read the rest

Benchmarks: TollBit (3,000+ sites, 1.5B quarterly bot scrapes monitored, company figures) and Cloudflare Pay Per Use (beta Sept 30 2026) target publishers and AI companies; Agent Door targets small merchants (restaurants, salons, shops) with a self-serve, per-agent allow/deny/price door. Revenue figures are hypotheses.

DescriptionHypothesis

Primary users are AI agents (via MCP/HTTP), paid for by merchants who self-serve. Product: a script or plugin (Shopify, WordPress, Webflow, plain site) that (1) identifies visiting agents, (2) lets the owner set allow/deny/price per agent and per action in a dashboard, (3) exposes a clean agent-readable surface (menu, …

Read the rest

hours, availability, booking request endpoint, quote) so agents stop scraping the human page, and (4) logs every agent action with receipts. Pricing: $19/month per site plus usage fees per accepted action (Hypothesis). Differences: TollBit is publisher/content-licensing focused with an enterprise flavor; site-compatibility trackers map which sites block which agents; Agent Handoff Desk (existing HBV entry) hands blocked tasks to humans. Companies already serving agents as primary users (watch list): TollBit (agent sites, bot paywall, MCP), Cloudflare Pay Per Use, Skyfire (agent payments/identity), NitroStack (hosted MCP servers, #1 on Product Hunt Oct 19 with 295 votes), Shopify's agent catalog access, Meta's Muse retailer program.

Target customer

Small merchants (restaurants, salons, local shops) and the agent builders whose agents visit them

Pain

Agents hit small sites without rules; owners can't allow, price or block them; agents scrape and fail

Why this score

Why it scored lower: Incumbents moving down-market; unproven merchant demand

MoatHypothesis

Plugin distribution + per-agent policy and receipts data. Installed base and a registry of agent identities/behavior; SEO for 'allow AI agents on my site'. Defensibility: moderate if distribution wins; status: Hypothesis.

Demand signals

X posts (7 days)

n/a

Signal trend

n/a

Web demandVerifiedSelf-reported

The Verge Oct 17 2026 (Verified); TollBit and Cloudflare traction (Self-reported/Verified); NitroStack PH #1 (Verified).

Market and competitors

TAMHypothesis

$2B (Hypothesis) — Small-site agent access, monitoring and transactions. Method: Hypothesis: ~20M small sites x ~$100/year long-term; not a sourced report.

SAMHypothesis

$100M (Hypothesis) — Shopify/WordPress/Webflow merchants with bookings or catalogs. Method: ~1M sites x ~$100/year.

SOMHypothesis

$114K ARR (Hypothesis) — 12-18 months. Method: ~500 sites x $19/month ≈ $114K ARR.

Competitors

TollBit, Cloudflare Pay Per Use, Skyfire, Shopify agent access, Cloudflare/Akamai bot tools, NitroStack

Business model

Monetization

$19/site/month subscription plus usage fee per accepted agent action

Revenue hypothesis (monthly)Hypothesis

Base ~$9.5K MRR, conservative ~$2K MRR at 12-18 months (Hypothesis)

Jobs to be done

When personal AI agents start visiting my site and trying to book or buy, I want to decide which agents may act, at what price, through a clean entrance, so I can get new customers from agents without chaos or abuse.

HXC

Dani · 38 · salon owner, Lisbon: Runs a salon with a Webflow site and Instagram; customers may soon ask their agents to book. Doesn't want bots scraping or double-booking; would pay $19/month for control and clean bookings.

Build and go-to-market

Tech stack

Next.js on Vercel; plugins (Shopify/WordPress/Webflow); MCP server; Supabase; Stripe; GitHub cesarx2 (one repo)

GTM channels

Plugin directories, docs SEO/GEO ('allow AI agents on my site'), MCP registries, integrations with agent builders, newsletter; no founder face, no paid ads

Risks

Main risk

Incumbents moving down-market; unproven merchant demand

All risks

Value Risk: Attention — Merchants may not feel agent demand yet. | Usability Risk: Supported — Allow/deny toggles and a simple menu. | Feasibility Risk: Attention — Agent identification is imperfect; policies change weekly. | Business Viability Risk: Risk — TollBit/Cloudflare/Shopify may cover small sites.

Next validation step

Interview 15 small merchants and 10 agent builders; publish a plugin waitlist with price

Sources

theverge.com https://www.akamai.com/newsroom/press-release/no-free-crawls-akamai-tollbit-and-skyfire-turn-traffic-into-revenue https://blog.cloudflare.com/pay-per-use/ https://docs.tollbit.com/docs/agent-site https://productlist.info/daily/2026-10-19

agent-servicebenchmark:TollBit, ai-agents, mcp, small-merchants, plugin, bet, hypothesis, idea, new

Want the next one the day it is scored?

The score, the sources, the risks and the next step, by email.

ProductBots.ai

Like this idea? Turn it into a plan with ProductBots.ai.

ProductBots.ai is an Agentic Product OS: agents that do the product work with you, from discovery to plan. Download the briefing for this idea and use it as the starting point.

Open ProductBots.ai

Preview the briefing

# Briefing: Agent Door

Source: Startup Ideas HQ (HB Ventures), found 2026-10-20. ProductBots Opportunity Score: 59/100, evidence confidence 41%.
Evidence labels (Verified, Self-reported, Hypothesis) are kept as written by the researcher.

## Goal for ProductBots.ai
Turn this idea into a validated MVP plan and a first build. Start from the next validation step; do not assume unlabelled numbers are verified.

## What it is
A one-snippet 'door' that lets a small merchant or site decide, per personal AI agent (Muse, Dots, Grok Bot, Instinct), whether to allow, block or charge it, and serves agents a clean machine-readable menu of bookable actions (availability, catalog, quotes).

BET / HYPOTHESIS. The Verge (Oct 17 2026): Yelp only allows agents that pay for data licensing, Delta blocks agent bookings, JPMorgan and Schwab block agents, while Meta partnered with Walmart, Gap and Sephora and Shopify opened its catalog to more agents. Big sites are choosing; small sites have no easy tool. Benchmarks: TollBit (3,000+ sites, 1.5B quarterly bot scrapes monitored, company figures) and Cloudflare Pay Per Use (beta Sept 30 2026) target publishers and AI companies; Agent Door targets small merchants (restaurants, salons, shops) with a self-serve, per-agent allow/deny/price door. Revenue figures are hypotheses.

## Customer and pain
Target customer: Small merchants (restaurants, salons, local shops) and the agent builders whose agents visit them
Pain: Agents hit small sites without rules; owners can't allow, price or block them; agents scrape and fail

## Market and competitors
TAM: $2B (Hypothesis) — Small-site agent access, monitoring and transactions. Method: Hypothesis: ~20M small sites x ~$100/year long-term; not a sourced report.
SAM: $100M (Hypothesis) — Shopify/WordPress/Webflow merchants with bookings or catalogs. Method: ~1M sites x ~$100/year.
SOM: $114K ARR (Hypothesis) — 12-18 months. Method: ~500 sites x $19/month ≈ $114K ARR.
Competitors: TollBit, Cloudflare Pay Per Use, Skyfire, Shopify agent access, Cloudflare/Akamai bot tools, NitroStack
Geography: Global except China; first US/EU/LatAm small merchants

## Business model
Monetization: $19/site/month subscription plus usage fee per accepted agent action
Revenue hypothesis (monthly): Base ~$9.5K MRR, conservative ~$2K MRR at 12-18 months (Hypothesis)

## Build and go-to-market
Build effort (agent-weeks): 3
Time to first revenue: 6-8 weeks
Tech stack: Next.js on Vercel; plugins (Shopify/WordPress/Webflow); MCP server; Supabase; Stripe; GitHub cesarx2 (one repo)
GTM channels: Plugin directories, docs SEO/GEO ('allow AI agents on my site'), MCP registries, integrations with agent builders, newsletter; no founder face, no paid ads

## Risks and next step
Main risk: Incumbents moving down-market; unproven merchant demand
Next validation action: Interview 15 small merchants and 10 agent builders; publish a plugin waitlist with price

## Sources
https://www.theverge.com/ai-artificial-intelligence/1013195/ai-agents-blocked-web https://www.akamai.com/newsroom/press-release/no-free-crawls-akamai-tollbit-and-skyfire-turn-traffic-into-revenue https://blog.cloudflare.com/pay-per-use/ https://docs.tollbit.com/docs/agent-site https://productlist.info/daily/2026-10-19

Existing businessAgent Services· 3 weeks ago
CreativeClaw-style copycat
CreativeClaw is a remote MCP server that gives Claude (and other agents) one creative studio across the top image, video, audio and 3D models; bootstrapped, one person, it did $100,077 in the last 30 days (Verified, Polar) after $5,254 in August, selling credits ($10 = 1,000) with no subscription.
62/100
confidence 55%
Why it scored lower: Margin and spike risk
Open ProductBots.ai →
New ideaAgent Services· 3 weeks ago
Brazil Data for Agents
A pay-per-call API and MCP server that gives AI agents clean, cached Brazilian data (CEP, CNPJ, holidays, FIPE) with agent-friendly schemas, so agents never scrape or guess.
61/100
confidence 28%
Why it scored lower: Agents may not pay
Open ProductBots.ai →
New ideaAgent Services· 3 weeks ago
Agent Handoff Desk
An agent-facing MCP/API where a personal or business AI agent that gets blocked on a website or needs a phone call hands the task to a vetted remote human operator, who completes it under the user's approval and returns structured proof — the digital 'last mile' counterpart to RentAHuman.
61/100
confidence 45%
Why it scored lower: Trust & safety incidents and RentAHuman/agent vendors moving into digital tasks
Open ProductBots.ai →

Research, not investment advice. Numbers keep the label they were published with.