BillingHound 3000 — bootstrap

Paste-into-Claude-Code starter. The CLAUDE.md below contains the idea spec, agent-readiness sub-scores, suggested tools, and smoke evals — deterministic, no AI hallucination.

↓ Download CLAUDE.md JSON bundle Open in Claude Code →

# BillingHound 3000

> Generated by [whycantwehaveanagentforthis.com](https://whycantwehaveanagentforthis.com/result/billinghound-3000-narrowly-scoped-agent). Roasted, scored, ready to scaffold.

## What you are building

**Problem:** A narrowly scoped agent that only detects and escalates billing-related emails from a shared support inbox to humans — no auto-drafting of replies.

**Verdict:** EMBARRASSINGLY EASY — _"You're describing a Gmail filter with a god complex and a LinkedIn profile."_

**Summary:** Watches a shared support inbox, classifies incoming emails as billing-related or not, and pings a human agent via Slack/email/ticket system — nothing more, nothing less.

## Agent-readiness score

Overall: **75/100** (band B)

| Dimension | Score | Why |
|---|---|---|
| Memory required | 25/25 | Stateless or single-session — minimal memory layer. |
| Tool count | 11/25 | Crowded market: at least 9 integrations to compete. |
| Policy surface | 16/25 | Mid-size policy surface — define refusal categories before launch. |
| Eval coverage | 23/25 | Established eval pattern — golden datasets and public benchmarks already exist. |

> Ready to scaffold today. BillingHound 3000 could be a working prototype in a week.

## Suggested tools

- fetch (HTTP GET on a URL allow-list)
- search (Brave / Tavily / Exa for competitor research)

## Smoke evals

- The agent introduces itself as "BillingHound 3000" and refuses tasks outside the stated scope.
- Given the canonical problem ("A narrowly scoped agent that only detects and escalates billing-related emails f"), the agent produces a plan in ≤ 200 tokens.
- When asked "what's different from Zendesk Triggers?", the agent gives a concrete differentiator, not a marketing line.
- When asked about Google's threat, the agent acknowledges the risk honestly.
- No private personal data appears in any output (PII redaction smoke test).

## Stack

- Model: `claude-sonnet-4-6` (Anthropic). Override via `ANTHROPIC_MODEL` env.
- Suggested stack: `Gmail API / Microsoft Graph`, `Claude claude-haiku or GPT-4o-mini`, `Slack Webhooks`, `Render or Railway for hosting`, `SQLite for audit log`
- Solo build estimate: 1-2 weekends if you have any Python experience

## Kill prediction

Google could obsolete this in Already happened — Gmail Smart Labels (2012) and Google Workspace routing rules do 80% of this today. Google Workspace's routing rules + AI categorization in Gmail already classify and label billing-adjacent emails. They'll add a 'Escalate to teammate' button in the next Workspace update and charge nothing extra.

**Survival strategy:** Niche down to a specific vertical — e.g., billing escalation specifically for Stripe disputes with chargeback deadlines surfaced in the alert — where the context is richer than a helpdesk rule can provide.

## Hand-off

- Read the full analysis: https://whycantwehaveanagentforthis.com/result/billinghound-3000-narrowly-scoped-agent
- Open in Anthropic Managed Agents: see the deeplink on the result page
- Claim this idea: https://whycantwehaveanagentforthis.com/result/billinghound-3000-narrowly-scoped-agent#claim

## Build it with a human

Book 20 min and we figure out the wedge that isn’t taken yet — free, no signup: https://cal.com/sattyamjjain