Metaprod: Spin Up a Product Manager Agent in Warp
Today I extended the Weather App PM test to have the product manager agent running directly in the command line and create a new experiment: A project called Metaprod.
Using Warp and the MCP server connected to Linear, I asked the agent to create a PRD (Product Requirements Document) and automatically generate tasks inside a new Linear project.
The flow was simple:
Prompt the agent with an idea.
Let it draft the PRD.
Push that PRD + tasks into Linear.
Watch as it scaffolds the project with epics, subtasks, estimated timelines, and success criteria.
The output was surprisingly thorough—it created a project foundation, architecture tasks, even a database setup. All of this was written into Linear without me needing to touch the UI.
What I really like about Warp is the clean and simple task interface. It feels natural to design, test, and automate workflows there. With auto-approve turned on, the process flowed smoothly without constant confirmations.
While the PRD was a bit more verbose than I might have wanted (next time I’ll specify something leaner), the experiment proved the concept:
Built by my PM Agent
👉 A product manager agent in your terminal.
Metaprod shows how we can outline and track projects directly from the command line, using LLMs to handle the scaffolding that product managers usually do manually.
This is just a first test, but the potential is clear—whether you’re spinning up new apps, agents, or workflows, having a product agent at your side means the groundwork gets laid out for you instantly.
—
Now on to updating tickets dynamically…
In my first video, I showed how an MCP server connected to Linear could create a PRD and scaffold tasks directly from the command line. This time, I wanted to see if the agent could go a step further: update existing tickets.
I pulled up one of the first epics, grabbed its ID, and ran an update command. The goal was to shift the plan toward a lean build approach with a specific stack:
Next.js for the frontend
Supabase for auth + database
Playwright for UI testing
Vercel for deployment
And keep the app lean: a simple landing page with login, a user dashboard, and profile settings (including Linear and OpenAI key management).
The agent not only updated the ticket but also refreshed subtasks like “Build foundation,” “Implement AI-powered PRD generation,” and “Set up the dev environment.” Even better—it adjusted the estimated timeline (cutting down from 46 weeks to 2–3) and deferred complex integrations to phase two.
This was a pleasant surprise. It showed that the MCP agent is not just about ticket creation—it can maintain and refine existing tickets in Linear as the project evolves.
