cd /news/ai-agents/why-i-open-sourced-the-tool-i-built-… · home topics ai-agents article
[ARTICLE · art-98528] src=dev.to ↗ pub= topic=ai-agents verified=true sentiment=↑ positive

Why I open-sourced the tool I built "just for myself"

A developer open-sourced DeepSeek Phone Harness, a tool originally built for personal use to remotely control a desktop agent from a phone. The project, released under MIT license with zero dependencies, aims to share hard-won knowledge about the agent's event stream, including a critical bug where the agent-question channel is separate from the approval channel, causing silent deadlocks. The developer found that open-sourcing improved the tool significantly within a week.

read2 min views1 publishedAug 16, 2026

It started as a selfish project: a phone remote for my own desktop agent. Three reasons made me push it public. The third one surprised me.

I built DeepSeek Phone Harness for one reason: I was in bed at 1 a.m., wanted to send a task to the agent on my computer, and refused to get up. Pure selfishness.

It worked so well that I kept using it. And then I had to decide: keep it private, or open it up?

The hardest bug — the one that made tasks hang forever — was invisible until you knew it existed: DSH's agent-question channel is separate from its approval channel. Miss it, and every ask_user_question

call deadlocks your pipeline silently.

There's no way to discover this by reading docs. You have to capture a real event stream, notice a frame type nobody told you about, and chase it for two days. That kind of hard-won knowledge should be public — someone else shouldn't have to re-pay that tuition.

"Phone as a remote for a desktop agent" is a genuinely useful pattern. But right now it lives in scattered, platform-specific implementations. An MIT-licensed, zero-dependency reference implementation lowers the floor: anyone can clone it, read ~4000 lines, and build their own — for DeepSeek Harness, or any agent runtime with a similar event model.

Open-sourcing it forced me to do things I wouldn't have bothered with otherwise:

The tool got better in the week after open-sourcing than in the week before. That surprised me, and it's the reason I'll keep building in public.

Nothing, really. But if you take one thing: when you fix a nasty bug nobody warned you about, write it down and put it where others can find it. That's the entire spirit of this project.

Repo (MIT, zero dependencies, three-minute setup): Have you open-sourced something you built "for yourself"? What changed? I'd genuinely like to know.

── more in #ai-agents 4 stories · sorted by recency
── more on @deepseek phone harness 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/why-i-open-sourced-t…] indexed:0 read:2min 2026-08-16 ·