New ideaAgent-services

Agent Cancel Desk

An MCP/API that personal AI agents call to actually cancel a user's subscription: it finds the official cancellation path, executes it with the user's explicit approval, and returns proof (protocol number or confirmation) — paid only on success.

3 weeks ago · Scored 21 Oct 2026 · ID HBV-261021-08

Working title: Agent Cancel Desk (working name)

Summary

BriefingVerifiedSelf-reportedHypothesis

New / agent-services idea (hunch). Benchmark: flightclaw (agent books and pays for flights end-to-end and shows a visible service fee in its demo — Self-reported demo, $25.65 fee shown) and Agentik (pay-per-call tool catalog for agents; 2,429 tools, 302,049-impression launch post — Verified X).…

Read the rest

Neither cancels subscriptions. Existence check: Searches on 21 Oct 2026 found: Kordi MCP (returns a cancellation deep link/route, does not execute), LemSubs agent API (manages subscription data; its docs say it cannot cancel external vendor subscriptions), Recharge MCP (merchant-side cancellation for Recharge stores only), and a hackathon repo using voice calls (Kaeru). No agent-callable service that executes consumer cancellations across vendors and returns proof was found. X: 1,540 posts/7d (Down 12% Thu→Tue (full UTC days: Thu 15 251 → Tue 20 221); peak Thu 15 251; Tue 20 is 12% below peak; window 2026-10-14 10:00 UTC → 2026-10-21 10:00 UTC (Wed 14 Oct and Wed 21 Oct are partial UTC days; full days Thu 15 → Tue 20)). Revenue hypothesis: ~$2K–$8K/month at month 9 (HYPOTHESIS: 1,500–4,000 confirmed cancellations × ~$1.50–$2 net).

DescriptionHypothesis

Read the rest

sistants, OpenClaw-type personal agents) acting for people who want a subscription gone; human beneficiaries are adults in Brazil first, then the US/Mexico. Pain: Cancelling is the step agents still cannot finish: existing agent tools return a link or a route, and the human still has to log in, find the hidden button or call support. People complain publicly every day that they cannot cancel. Monetization: Success fee per confirmed cancellation paid from the agent's wallet (hypothesis $1–$3); free route lookup; listing on agent tool marketplaces (Agentik-type) as distribution. stack: MCP server + Next.js approval page on Vercel; Postgres route DB; x402 or prepaid wallet billing. GTM: List on MCP registries and agent tool catalogs; SEO/GEO pages "how to cancel <service> in Brazil" that agents and humans cite; faceless how-to clips.

Target customer

Primary user: consumer AI agents (ChatGPT, Claude, Meta Muse-style assistants, OpenClaw-type personal agents) acting for people who want a subscription gone; human beneficiaries are adults in Brazil first, then the US/Mexico.

Pain

Cancelling is the step agents still cannot finish: existing agent tools return a link or a route, and the human still has to log in, find the hidden button or call support. People complain publicly every day that they cannot cancel.

Why this score

Why it scored lower: Vendors block automation or require phone calls; success rate may be too low to earn a per-success fee.

MoatHypothesis

Route DB + success data: Fresh, tested cancellation routes per vendor and country are hard to keep current. Every success/failure improves routing; agent platforms prefer the tool that finishes the job. Status: Hypothesis.

Demand signals

X posts (7 days)

1540

Signal trend

Down 12% Thu→Tue (full UTC days: Thu 15 251 → Tue 20 221); peak Thu 15 251; Tue 20 is 12% below peak

Web demandVerified

Cancellation pain: 1,540 posts/7d in EN+PT, flat (Verified X). Agent-capability pain: 1,172/7d (Verified X). Search volume for "how to cancel X" not measured this run. Other X counts (same window, Verified X API): 1,172 posts: ("for my agent" OR "agent needs" OR "agents need" OR "agent can't") (pay OR book OR buy OR login OR password OR phone) lang:en

Market and competitors

TAM

Not sized this run — Scope: Consumer subscription management. Method: No market report fetched. (Basis: n/a)

SAMHypothesis

~$100K–$500K / yr (hypothesis) — Scope: Agent-initiated cancellations BR+US. Method: HBV assumption. (Basis: Hypothesis)

SOMHypothesis

~$24K–$96K ARR (hypothesis) — Scope: Month-9 run-rate. Method: From rev ×12. (Basis: Low)

Competitors

Kordi (routes), LemSubs (subscription data), Rocket Money and similar apps (human-facing concierge, not agent APIs), consumidor.gov.br (Brazil complaint channel). No direct agent-executing competitor found.

Business model

MonetizationHypothesis

Success fee per confirmed cancellation paid from the agent's wallet (hypothesis $1–$3); free route lookup; listing on agent tool marketplaces (Agentik-type) as distribution.

Revenue hypothesis (monthly)Hypothesis

~$2K–$8K/month at month 9 (HYPOTHESIS: 1,500–4,000 confirmed cancellations × ~$1.50–$2 net).

Jobs to be done

When my user told me to cancel a subscription, I want to execute the cancellation with one approval and get proof, so I can report back done instead of handing the user a link.

HXC

Agent of Paula, 38, Recife (illustrative) · Paula asks her assistant to "cut my subscriptions"; it can list them but stalls at cancelling. Buying signal: Her agent calls Cancel Desk, she taps approve, and gets a protocol number — the agent pays R$8 from its balance.

Build and go-to-market

Tech stack

MCP server + Next.js approval page on Vercel; Postgres route DB; x402 or prepaid wallet billing

GTM channels

List on MCP registries and agent tool catalogs; SEO/GEO pages "how to cancel <service> in Brazil" that agents and humans cite; faceless how-to clips

Risks

Main risk

Vendors block automation or require phone calls; success rate may be too low to earn a per-success fee.

All risksVerified

Value: Supported — Pain Verified daily | Usability: Attention — User must trust agent to act | Feasibility: High — Vendors resist automation | Business Viability: Unvalidated — Per-success WTP by agents

Next validation step

Build routes + execution for 20 top Brazilian subscriptions; kill if fewer than 60% of 100 test cancellations succeed end-to-end or fewer than 3 agent platforms list it in 30 days.

Sources

kordiapp.com (via github.com/kordi-labs/kordi-mcp-server)
lemsubs.com
agentik.cc
X API counts

ideaagent-servicesmcpsubscriptionscancellationsuccess-feebrazilusbenchmark-flightclawbenchmark-agentik

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 Cancel Desk (working name)

Source: Startup Ideas HQ (HB Ventures), found 2026-10-21. ProductBots Opportunity Score: 61/100, evidence confidence 45%.
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
An MCP/API that personal AI agents call to actually cancel a user's subscription: it finds the official cancellation path, executes it with the user's explicit approval, and returns proof (protocol number or confirmation) — paid only on success.

New / agent-services idea (hunch). Benchmark: flightclaw (agent books and pays for flights end-to-end and shows a visible service fee in its demo — Self-reported demo, $25.65 fee shown) and Agentik (pay-per-call tool catalog for agents; 2,429 tools, 302,049-impression launch post — Verified X). Neither cancels subscriptions. Existence check: Searches on 21 Oct 2026 found: Kordi MCP (returns a cancellation deep link/route, does not execute), LemSubs agent API (manages subscription data; its docs say it cannot cancel external vendor subscriptions), Recharge MCP (merchant-side cancellation for Recharge stores only), and a hackathon repo using voice calls (Kaeru). No agent-callable service that executes consumer cancellations across vendors and returns proof was found. X: 1,540 posts/7d (Down 12% Thu→Tue (full UTC days: Thu 15 251 → Tue 20 221); peak Thu 15 251; Tue 20 is 12% below peak; window 2026-10-14 10:00 UTC → 2026-10-21 10:00 UTC (Wed 14 Oct and Wed 21 Oct are partial UTC days; full days Thu 15 → Tue 20)). Revenue hypothesis: ~$2K–$8K/month at month 9 (HYPOTHESIS: 1,500–4,000 confirmed cancellations × ~$1.50–$2 net).

## Customer and pain
Target customer: Primary user: consumer AI agents (ChatGPT, Claude, Meta Muse-style assistants, OpenClaw-type personal agents) acting for people who want a subscription gone; human beneficiaries are adults in Brazil first, then the US/Mexico.
Pain: Cancelling is the step agents still cannot finish: existing agent tools return a link or a route, and the human still has to log in, find the hidden button or call support. People complain publicly every day that they cannot cancel.

## Market and competitors
TAM: Not sized this run — Scope: Consumer subscription management. Method: No market report fetched. (Basis: n/a)
SAM: ~$100K–$500K / yr (hypothesis) — Scope: Agent-initiated cancellations BR+US. Method: HBV assumption. (Basis: Hypothesis)
SOM: ~$24K–$96K ARR (hypothesis) — Scope: Month-9 run-rate. Method: From rev ×12. (Basis: Low)
Competitors: Kordi (routes), LemSubs (subscription data), Rocket Money and similar apps (human-facing concierge, not agent APIs), consumidor.gov.br (Brazil complaint channel). No direct agent-executing competitor found.
Geography: Brazil first (PT), then US and Mexico

## Business model
Monetization: Success fee per confirmed cancellation paid from the agent's wallet (hypothesis $1–$3); free route lookup; listing on agent tool marketplaces (Agentik-type) as distribution.
Revenue hypothesis (monthly): ~$2K–$8K/month at month 9 (HYPOTHESIS: 1,500–4,000 confirmed cancellations × ~$1.50–$2 net).

## Build and go-to-market
Build effort (agent-weeks): unknown
Time to first revenue: ~3–5 weeks after launch
Tech stack: MCP server + Next.js approval page on Vercel; Postgres route DB; x402 or prepaid wallet billing
GTM channels: List on MCP registries and agent tool catalogs; SEO/GEO pages "how to cancel <service> in Brazil" that agents and humans cite; faceless how-to clips

## Risks and next step
Main risk: Vendors block automation or require phone calls; success rate may be too low to earn a per-success fee.
Next validation action: Build routes + execution for 20 top Brazilian subscriptions; kill if fewer than 60% of 100 test cancellations succeed end-to-end or fewer than 3 agent platforms list it in 30 days.

## Sources
https://kordiapp.com/mcp (via github.com/kordi-labs/kordi-mcp-server); https://lemsubs.com/agent-api; https://www.agentik.cc; X API counts

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