beafk - Run Claude Code, Codex, Grok and Kimi as one coding crew

by
beafk runs Claude Code, Codex, Grok Build and Kimi Code on your own machines and lets them work together on one job. A lead agent delegates work while you steer the crew from your phone. Teams get roles, per-machine access grants and a signed audit trail, with pricing per machine — not per seat. Connect directly over Tailscale, so your code, chats and agent credentials never pass through beafk.

Add a comment

Replies

Best
Maker
📌
Hey Product Hunt 👋 I started building beafk because I realized something slightly ridiculous about the way I was working with coding agents. Most of my development day had become launching several agent sessions, waiting for them to finish, answering the occasional question, and spending the time in between staring at the monitor. Not long ago my son was born, and I wanted to spend more time with him, especially outside the house. That made me ask a pretty simple question: What is the point of having AI agents do the work if I still can't walk away from them? I tried Cursor on mobile, but I never really liked the experience. The cost of running agents also felt completely unpredictable, which I think will sound familiar to a lot of people. What I did love about Cursor was how easy it was to connect different providers and orchestrate them inside a single session. But the bill I got from them last month was the final push for me to look for another way. Around that time K3 came out, and I decided to try their subscription first. That was when I realized just how much I could actually get for $200. After that I subscribed to Claude Code and Grok too. I already had Codex :) At first I ran all of them from my phone through Termius. That is when something really clicked for me. The native CLIs worked noticeably better with their own models. The quality of the results was better, and I could use the subscriptions I was already paying for instead of thinking about the cost of every individual run. So now I had most of what I wanted: predictable subscriptions, native coding agents, and several different providers I could choose from depending on the task. What I did not have was a good way to use them together from my phone. Using a terminal on a phone works, but for obvious reasons it is not exactly the experience I wanted. I wanted to orchestrate different agents inside one session. I wanted to use specific models for specific jobs, delegate parts of a task without overloading the main agent's context, and get complex work done faster, cheaper and with better results, without giving up the native CLIs. I also wanted things like MCP servers and skills to be available across different CLIs automatically, instead of configuring the same setup again and again. On July 22 I ran the first few sessions and made the decision to stop working on my previous product and focus on beafk. I used Termius on my phone to build only the very first version of the web interface. Everything after that was built inside beafk itself. The next 37 days of development happened from my phone. I've spent more time with my family, made it to the gym much more often, and generally stopped structuring my day around being close to my desk. It was a wonderful month. And for me, that became the strongest validation of the idea. I was building a product because I wanted to be AFK, and then I actually got to be AFK while building it. I spent a lot of time trying to make the mobile interface feel like something that was actually designed for a phone, not a desktop product squeezed onto a smaller screen. There are quick tabs, flexible controls, very fast synchronization between devices, and a bunch of small details I care about probably more than I should. I especially hope you notice the way you can collapse and expand the conversation while working :) There is another thing I cared about a lot. Your code does not go through beafk. Your phone connects directly to your machine over your own Tailscale network. beafk issues the short lived certificate that opens the door, and then we are off the wire. Your code, chats and agent credentials stay on your machine. I also wanted beafk to work for more than one developer. You can invite a whole team, assign admins and members, control exactly who can access each machine, and review the certificates issued by the workspace in a signed, hash chained audit log. And the pricing still follows machines, not seats. I have spent a lot of time testing all of this, but I am sure there are still rough edges. This is the point where I really need people other than me to use it and tell me what is confusing, what breaks, what feels unnecessary, and what you wish worked differently. beafk has a 7 day free trial with no credit card required. And if seven days is not enough to properly test it, message me during launch week and mention Product Hunt. I’ll be happy to extend your trial! More than anything, I am looking for feedback right now. Thank you for checking out beafk ❤️