Replywise
Draft customer replies that still sound like your team.起草仍像你团队口吻的客户回复。
Financial summary财务摘要
Asking multiple: ≈2.1× trailing annual profit. Trailing annual profit ≈ $32,400.要价倍数：约等于近一年利润的 2.1 倍。近一年利润约 $32,400。
Cost breakdown成本明细
| LLM API (OpenAI) | $820 |
|---|---|
| Hosting (Vercel + Railway) | $95 |
| Vector DB | $60 |
| Stripe | $140 |
| $35 | |
| Monitoring | $50 |
| Misc SaaS | $100 |
| Total monthly costs | $1,300 |
Implied monthly profit $4,200 − $1,300 = $2,900 (listing shows rounded owner profit of $2,700).隐含月利润 $4,200 − $1,300 = $2,900（挂牌展示圆整后业主利润 $2,700）。
Owner workload业主工作量
Heavier than other examples: ~5 hours on support and prompt tuning, ~2 hours on infra/cost watch, ~1 hour on onboarding calls for larger teams (optional). Model price changes need monitoring.比其他示例更重：约 5 小时支持与提示调优，约 2 小时基础设施/成本盯盘，约 1 小时大客户可选入驻通话。需关注模型价格变化。
Tech stack技术栈
- Next.js
- TypeScript
- Postgres
- pgvector
- OpenAI API
- Stripe
- Vercel
- Railway
Customers & retention客户与留存
186 workspaces. Top customer = 9% of MRR (usage-based overage). Gross margin after LLM ≈64%. Monthly logo churn ~5%. Expansion from seat adds is healthy. Customers: Shopify brands and B2B SaaS support teams.186 个工作区。最大客户占 MRR 9%（用量超额）。扣除 LLM 后毛利率约 64%。月 logo 流失约 5%。席位扩张健康。客户：Shopify 品牌与 B2B SaaS 支持团队。
No fine-tuned custom model — all prompts + retrieval. Fine-tuning could be upside for a technical buyer.无自研微调模型——全靠提示词与检索。对技术型买家，微调可能是上行空间。
Handover checklist交接清单
- API keys rotated; usage caps documentedAPI 密钥轮换；用量上限已文档化
- Prompt library + eval set (golden replies)提示词库 + 评测集（黄金回复）
- 30-day transition with weekly cost review30 天过渡，每周复盘成本
- SOC2 not started — buyer should plan if selling to enterprise未启动 SOC2 — 若面向企业需买家自行规划
- Intercom replacement is native inbox — no Intercom dependency原生收件箱，不依赖 Intercom
Seller story卖家故事
Replywise grew from a side project helping our own support queue. AI costs are the main variable. I’m selling because a conflicting full-time offer starts soon and I won’t half-maintain an AI product. Buyer should be comfortable watching token spend.Replywise 源于帮助自己支持队列的副业。AI 成本是主要变量。出售是因为即将开始有冲突的全职工作，不愿半吊子维护 AI 产品。买家需能盯住 token 支出。