Should My AI Coding Agent be Open Source?
About this lesson
Join the community: https://mrc.fm/cmc I built Tank to run my own AI coding agents and it has taken on a life of its own. So today I am going live to clear up what Tank actually is, and to talk about something a lot of you have been asking for. Quick version: Tank is mission control for AI coding agents. You give an agent a task, it runs in a real terminal you can watch and jump into, with live feedback, a build queue for overnight work, and scheduled jobs. A human stays in the loop the whole way. It is not auto merge and it is not a black box. It is your command centre for Claude Code, Codex, Grok and more. Now the big one. A lot of you have asked me to open it up so you can run the code yourself. I have been seriously exploring exactly that: an open, source available core you could try, self host and build on, with the advanced features as optional modules. Today I want to tease where this is heading and hear from you. So come and tell me: would you run Tank yourself? Would you build a module? Should the core be open? Bring your questions, I am answering live. Get early access and help shape what happens next: https://mrc.fm/cmc 00:00 Intro: What is Tank? 01:03 The Open Source Question 01:58 When AI Agents Take Over Your Socials 03:52 Tour of the Tank UI 04:14 What Problem Does Tank Solve? 07:08 Live Demo: Spawning Tasks & Servers 11:04 Clarifying the Focus (Tank vs. Auto PR Mergers) 12:10 How Tank Differs from Cursor & Hermes 13:44 Managing API Costs & Rotating AI Models 14:55 Why a Product Model Over Community-Only? 16:35 Future Vision: Plugins & Marketplace Add-Ons 22:47 Multi-Agent Architecture to Prevent Bugs 25:52 Discussing the "Open Core" Model 30:10 Managing Rapid Community Growth 32:21 Introducing the Core Team & Outro
DeepCamp AI