Fomi 是一款 macOS 专注工具，于 2026 年 2 月 19 日（上午 5:30）被评测，尝试解决一种常见的阻碍失效问题：同一个网站会因任务情境不同而同时可能有用也可能分心。它不会封锁固定的 App 或网域，而是先询问你正在做什么，接著跨 App 监测萤幕上的活动，并在情境中对行为进行分类。该产品提供 3-day 免费试用，之后每月费用为 $8，将自己定位为低成本订阅制的生产力助理。
在约 2 天的测试期间，该应用程式使用了 3 种状态的注意力系统，配有绿／黄／红指示与递进式回馈，并在侦测到分心时显示红色画面的番茄警示。回报显示其在情境区分上的表现强劲，例如可在同一平台上允许相关阅读、同时标记不相关浏览，并且也能区分在 Reddit 或 YouTube 上的生产性与非生产性使用。评测者指出有一些误报，但可透过「误报」动作修正，而团队表示模型准确率正持续迭代提升。
主要的权衡在于隐私：Fomi 会频繁撷取萤幕截图，并将已遮罩的影像传送至 OpenAI GPT-5 Mini 进行一次性分析；同时该公司声称其端不做持久化伺服器储存，且仅以 RAM 处理。即使已在本地对姓名、电话号码、电子邮件、密码与卡号进行遮罩，观察到的上传量在一般工作日仍约为 0.5 GB，显示资料吞吐量很高。实务上的含意是：此工具短期内可能改善专注习惯，但由于持续萤幕资料传输的规模与敏感性，对于要求严格机密性的工作流程并不适合。
Fomi is a macOS focus tool reviewed on February 19, 2026 (5:30 AM) that tries to solve a common blocker failure: the same site can be both useful and distracting depending on task context. Instead of blocking fixed apps or domains, it asks what you are working on, monitors on-screen activity across apps, and classifies behavior in context. The product offers a 3-day free trial and then costs $8 per month, positioning itself as a low-cost subscription productivity assistant.
During testing over about 2 days, the app used a 3-state attention system with a green/yellow/red indicator and escalating feedback, including a red-screen tomato warning when distraction was detected. Reported performance was strong in contextual distinctions, such as allowing relevant reading while flagging unrelated browsing on the same platform, and similarly separating productive versus unproductive Reddit or YouTube use. The reviewer noted some false positives, but these could be corrected with a False Alert action, and the team says model accuracy is being iteratively improved.
The main tradeoff is privacy: Fomi frequently captures screenshots and sends redacted images to OpenAI GPT-5 Mini for one-time analysis, while the company claims no persistent server storage and RAM-only handling on its side. Even with local redaction of names, phone numbers, emails, passwords, and card numbers, observed upload volume was about 0.5 GB in a regular workday, indicating high data throughput. Practical implication: the tool may improve focus habits short term, but it is a poor fit for workflows requiring strict confidentiality due to the scale and sensitivity of continuous screen data transfer.