我创建了能够独立完成单人创始人讨厌的繁琐工作的人工智能代理。
嘿,HN,
我是一个独立创始人。每天早上我花两个小时做同样的事情:检查竞争对手、阅读人工智能新闻、监控我的Stripe仪表板、查看Google趋势以获取内容创意。
因此,我构建了Seleci来自动化这一切。
它的功能:你选择一个模板(市场脉搏、收入雷达、趋势侦察),点击部署,安排时间,然后每天早上你就会收到一份干净的Markdown报告——无需编码,无需Zapier流程,无需提示工程。
它的工作原理:
- 使用FastAPI后端,配备代理循环(工具调用LLM与网络搜索、Stripe API、Google趋势)
- 每个工具的速率限制,以防止代理循环失控(网络搜索每次运行限制为3次调用)
- React/Vite前端,Supabase身份验证,部署在Koyeb和Vercel上
- 代理实际上执行工具并返回结构化结果——而不是一个聊天机器人包装
它不是:
- 另一个ChatGPT的皮肤
- 无代码工作流构建器(没有节点,没有拖放)
- 试图取代开发者——这是为没有开发者的创始人准备的
现场演示:https://seleci.com
我正在用这个申请YC。希望通过Discord获得HN的严厉反馈——缺少什么,哪里有问题,什么会让你真正使用这个?
查看原文
Hey HN,<p>I'm a solo founder. Every morning I was spending 2 hours doing the same things: checking competitors, reading AI news, monitoring my Stripe dashboard, looking at Google Trends for content ideas.<p>So I built Seleci to automate all of it.<p>What it does: You pick a template (Market Pulse, Revenue Radar, Trend Scout), click deploy, schedule it, and every morning you get a clean markdown report — no code, no Zapier flows, no prompt engineering.<p>How it works under the hood:<p>FastAPI backend with an agentic loop (tool-calling LLM with web search, Stripe API, Google Trends)
Per-tool rate limiting to prevent runaway agent loops (web_search capped at 3 calls/run)
React/Vite frontend, Supabase auth, deployed on Koyeb + Vercel
The agent actually executes tools and returns structured results — not a chatbot wrapper
What it's NOT:<p>Not another ChatGPT skin
Not a no-code workflow builder (no nodes, no drag-and-drop)
Not trying to replace developers — it's for the founder who doesn't have one
Live demo: https://seleci.com<p>I'm applying to YC with this. Would love brutal feedback from HN via Discord — what's missing, what's broken, what would make you actually use this?