You know that feeling when you have 40 tasks and no idea where you left off? You check your project board. Phase 3 is still “In Progress” from yesterday. Your teammate says they’re working on the backend, but the status says “Todo.” Somewhere between Slack, Git, and Linear, the truth got lost.
That was me three weeks ago. I was drowning in coordination overhead. Then I tried something stupid: I spent an hour writing rules for my AI. Now I don’t manage tasks anymore. The AI does.
The real productivity multiplier isn’t AI writing code—it’s AI managing the workflow. Most people obsess over AI code generation. They miss the bottleneck: coordination. Figuring out what to do next, syncing progress, tracking who owns what. That’s where the time goes.
Here’s the stack I used: MCP + Rules. MCP is a bridge protocol that lets your AI IDE talk to external tools. Think of it as a universal translator. You tell your AI “create a project in Linear” and it just does it—no API keys, no manual clicks. Linear has a first-party MCP connector. I configured it in Trae (a code editor) in under five minutes.
Then came the Rules file. A simple markdown document that tells the AI how to behave. I didn’t write it from scratch. I fed the MCP docs and Trae’s Rules spec to the AI itself and said: “Write me a rule that says: every time I say ‘start working on [task]’, you mark that task as ‘In Progress’ in Linear, create a Git branch, and begin coding.” It wrote the rule. I tested it. It worked.
The paradox: to achieve zero-effort AI management, you must first invest significant effort in crafting precise rules. That upfront investment is what most people skip. They want magic. I wanted a system that could run itself after I set the direction.
The result: I now have 40 tasks in a Linear project. When I say “Start Phase 1, build frontend and backend,” the AI automatically marks Phase 1 as In Progress, starts coding, and when I finish, it moves it to Done and marks Phase 2 as In Progress. I never open Linear. I never update a status. I just talk to my AI.
Team collaboration? Even better. Two developers, each with their own branches. The AI syncs everything to one board. Everyone sees the same truth. No more “did you finish that?” messages.
You don’t need to manage your AI. You need to teach it to manage itself. The moment you stop treating AI as a code generator and start treating it as a project manager, everything changes.
So here’s the twist: I thought I was building a smarter coder. I was actually building a better to-do list—one that runs itself. The bottleneck wasn’t code. It was coordination. And now that coordination is automated, I can focus on the only thing that matters: deciding what to build next.
FAQ
Q: Doesn't this require a lot of upfront setup and customization?
A: Yes, it does. You need to configure MCP (a few minutes) and write a Rules file (an hour or two). But that investment pays off immediately—you stop context-switching between tools and start working in a single flow. The alternative is wasting hours every week on manual status updates.
Q: What's the practical first step for someone who wants to try this?
A: Pick a project management tool that supports MCP—Linear is a great start. Install the MCP connector in your IDE (Trae, VS Code, etc.). Then ask your AI to write a basic Rules file: 'When I say start working on X, update Linear status to In Progress and create a Git branch.' Test it on one task, then iterate.
Q: Isn't this just automating busywork that could be done with a script?
A: No, because the key insight is the natural language interface. You don't write a script for every scenario. You let the AI interpret your intent and execute the right actions. The rules provide guardrails, but the AI handles the context. This is a step beyond automation—it's autonomous coordination.