reddit · SaaS · 10-21 14:50 59 分 ← 返回列表

I built an agent builder for developers. The people who wanted it weren't developers.

v1 was a developer tool. You describe an agent, it writes a LangGraph.js agent, checks it compiles, and you export the code. I was proud of that part. Then I talked to people: - developers would rather write it themselves... fair - the ones who actually wanted it were founders, ops people and agencies doing the same tasks by hand every day - they didn't care about the code. they cared that it runs inside Slack & Gmail and asks before it sends anything So I changed th

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：非技术的创始人、运营和代理机构每天手工重复同样的任务，他们想要的是“描述任务就能跑起来”的现成 agent，而不是代码；并且这些 agent 必须能跑在 Slack/Gmail 里，发送前还要先问人。

目标用户：中小企业创始人、运营负责人、营销/外包代理机构等非技术但高频重复办公任务的团队

证据：the ones who actually wanted it were founders, ops people and agencies doing the same tasks by hand every day / they didn't care about the code. they cared that it runs inside Slack & Gmail and asks before it sends anything / Now it's getting traction

变现：订阅制 SaaS：按 agent 数量或月度执行次数分档（约 $29–99/月），代理机构按席位/客户数加价；可加“审批日志与合规”高级版

MVP：两周版：只接 Slack + Gmail 两个连接器，任务用自然语言描述即生成 agent 流程；核心差异点是“发送前必须点击审批”（Slack 按钮确认）；内置 5–8 个高频模板（客服回复草稿、线索跟进、每日摘要、发票催收）；提供执行日志。不做代码导出、不做通用编排画布。

竞品：Lindy、Zapier Agents、Gumloop、Relevance AI、n8n、Relay.app

摘要：创始人把 v1 做成给开发者的「生成可编译 LangGraph.js agent 代码」工具，但开发者更愿意自己写；真正想要的是创始人、运营和代理机构，他们不关心代码，只关心 agent 能跑在 Slack/Gmail 里且发送前先征求确认。改用「描述任务即得可用 agent」的定位后开始获得 traction。这说明付费方与产品亮点常常不一致，机会在于面向非技术运营人员的、带人工审批护栏的轻量 agent 工具。