PrumoGrid

A WORKSPACE FOR PEOPLE WHO BUILD SOFTWARE

Multiple terminals.
One project in order.

Your command-line agents, terminals and projects in one window. Follow the work, continue conversations and see what needs your attention.

Download for WindowsExplore the product

Windows 10 / 11 · x64/v0.5.11 · Pre-alpha

Runs on your computerUses the CLIs you already haveNo PrumoGrid account

BUILT FROM REAL DEVELOPMENT WORK

Opening another agent is easy.
Keeping the work aligned is harder.

The API is in one terminal. The frontend is in another. One agent implements, another reviews. You repeat the same specification, search for the last response and try to remember who is waiting for whom. PrumoGrid was built to organize this part of AI-assisted development.

More context in the project. Less work passing context between windows.

Meet the creator and read the story ↗

VIDEOS

Watch it in 1 minute.

What PrumoGrid is. Projects, terminals and CLI agents on Windows, the service that keeps everything running, where the name comes from and the areas of the screen.

See all 12 videos ↗

01 / INSIDE THE APP

Your terminal is still yours.

PrumoGrid adds space and context to the work you already do in the command line. Start with a grid. Add structure when you need it.

PRUMOGRID / WINDOWS
The API, the website and payments in one workspace.Actual 0.5.11 interface. Demo projects.
01 /

Less window switching.

Follow multiple projects, each with its own folder and terminals. Filter one project or see the whole workspace in All projects. Keep API, frontend and infrastructure work organized.

02 /

Pick up where you left off.

In persistent mode, terminals stay active when you close the window. After restarting Windows, resume identified Claude Code and Codex conversations through their CLIs.

03 /

Know what was verified.

An agent response and a passing command are different outcomes. Define a verification criterion and follow its result in the workflow.

04 /

Give each agent context.

Organize rules, roles and routines. Separate work using worktrees, and follow history and usage reported by supported CLIs.

02 / FROM PROMPT TO RESULT

Your way of working.
Room to grow.

You do not need an automation to open a terminal. Move from a simple session to multi-agent workflows at your own pace.

01

Open your project

Add your working folder. Terminals open in that context.

02

Choose your terminal

Use PowerShell, CMD, WSL or Git Bash. Start an installed, authenticated CLI such as Claude Code or Codex.

03

Follow up and continue

Work directly in the terminal or create steps. Check pending work, verify results and continue in the same session.

AN EXAMPLE WORKFLOW

  1. ImplementA task for your agent
  2. VerifyYour test command
  3. ReviewThe next step

You define the steps and criteria. PrumoGrid tracks execution.

IN PRACTICE

A workflow you can follow.

An example: implement a change, verify the result and hand it over for review. See where each part happens in the app.

01 /

Describe what needs to be done.

In New stage, write the task and check its project and agent. When the result needs to pass a test, add a verification command under More options.

  1. Click + New stage.
  2. Write the request and check its executor.
  3. Under More options, define a verification command when appropriate.

↳ The request is connected to the project, terminal and completion criterion.

Actual 0.5.11 interface. Demo task.
02 /

Stop repeating the same specification.

Define shared instructions and choose their scope: one project, all projects or a role. PrumoGrid organizes content in the files supported CLIs read, such as AGENTS.md and CLAUDE.md.

  1. Open Tools → Rules and roles… and create a new rule.
  2. Write the instruction and choose its scope.
  3. Check delivery and whether the agent needs to read it again.

↳ A common foundation for different agents, with room for specific instructions.

Actual 0.5.11 interface. Demo rule.
03 /

Hand work over to the next step.

Turn a recurring sequence into a routine. One step implements, another reviews and another tests. Choose executors and follow what is running, completed or waiting on a dependency.

  1. Open Routines, next to Workflow.
  2. Choose a routine and write this time’s task.
  3. Review the project and executors before starting.

↳ The sequence stays visible. You do not have to rebuild it for every task.

Actual 0.5.11 interface. Routine catalog.
04 /

The next task can use the context you already built.

With an open conversation, use Continue in this session to create another step in the same terminal. The conversation stays in the CLI; model and permissions follow the existing session.

  1. Check the previous step’s result.
  2. Choose Continue in this session.
  3. Write the next instruction and confirm.

↳ Continue your work with the context you already have.

Actual 0.5.11 interface. Demo session.

YOUR TERMINAL, YOUR TOOLS

Your harness still
has a place here.

Already have scripts, versioned prompts or your own layer for calling agents? Run your commands in PrumoGrid terminals. Use project organization and add rules and workflows as you need them.

Running your own command remains available. Features such as resuming, notifications and metrics depend on each CLI integration.

CLI options are configurable; empty fields preserve their defaults.

LEARNING PORTAL

From your first terminal
to multi-agent workflows.

Principles, one page per feature with step-by-step guides and current screenshots, complete recipes, troubleshooting and a glossary.

Open the learning portal

EVERYDAY FEATURES

EXTEND IT YOURSELF

Your workflow.
Your own plugins.

Create commands and open project terminals with a plugin built in your own repository. The guide includes complete files, a standalone packager and an installable example.

Open the plugin guide ↗

No pull request or access to PrumoGrid source code needed.

WHERE PRUMOGRID STARTED

Orchestrating multiple projects and coding agents

By Carlos Alvarenga

About the author

I’m a software architect with 30 years of experience in enterprise systems, including 18 years at MXM Sistemas, where I still work.

I hold an MBA in Artificial Intelligence and Big Data from USP/ICMC, with research into standardized code generation. My focus is applying AI to enterprise systems with governance, security and tangible results, preserving what already works.

I want to make technology more accessible, help train new developers and turn ideas into working solutions.

I tried several tools, but something was always missing.

I usually split a project into smaller projects, and I often work on several projects at once.

I wanted a single place to interact with multiple terminals, orchestrate coding agents in stages, get notified when something needed my attention and track what had already been done. All in one tool.

PrumoGrid grew out of that need, initially as a tool for my own work. Gradually, I turned those needs into specifications, validating and testing ways to bring terminals, projects, agents and pending work into the same workspace. It took about three months of work in my spare time to reach this pre-alpha version.

HOW IT WAS BUILT

PrumoGrid was built inside PrumoGrid.

The question I heard most was whether it had been made 100% by AI. It was not. The best way to explain it is through what I call the product curve, which covers concept, prototyping, validation, building, testing and delivery. There is no silver bullet along that curve. At each stage you apply the tools and techniques that fit your process and deliver results.

For PrumoGrid, concept, prototyping and validation started with an architecture-focused technical specification and clickable prototypes, until I had settled on an experience and workflow that made sense to me.

The building itself was done by orchestrating coding agents. First, PrumoGrid’s basic pattern was built, one project and several terminals. In one of them, Claude Code (Claude Opus 5) drove the implementation. In another, Codex (GPT-5.6 Sol) supported code review and usability testing. The final review, testing and delivery stayed with me.

I used frontier models because it was a large project, built outside working hours, and I wanted to spend my time deciding, not fixing code.

That is why I describe it as a three-way collaboration. AI wrote a good part of the code, but what the product is, its architecture, its patterns, how it works and when it is ready were my decisions. The result went beyond a prototype or an idea. It is a real coding-agent orchestration product that was used to build itself.

Carlos Alvarenga

03 / TRY IT

Give your project
a place to work.

Install the app, open your folder and start with a terminal. The installer includes what PrumoGrid needs to run.

v0.5.11 / Pre-alpha
PrumoGrid
Platform
Windows · x64
Size
90.4 MiB
Prepared
Download for Windows

Close the PrumoGrid window before installing; your terminals keep running in the service.

Free Use 1.0

This is an experimental release and the installer is not digitally signed yet. Windows may show an unknown publisher warning.

Before you install

  • Windows 10 1809+ or Windows 11, 64-bit.
  • Node.js is not required to run the app.
  • For agents, install and authenticate their CLIs. Provider accounts and costs are separate.
  • Git is required for code review and worktree features.

In this release / 0.5.11

  • Signed updates: PrumoGrid tells you when a new version is available, checks its signature and installs it once you confirm.
  • New step: task, agent and terminal in a single dialog. Workflows chain steps across agents and across projects.
  • Routines: three ready-made processes, plus routines saved from a workflow or built from scratch.
  • Attention gathers reviews and agent permission requests, including those from other projects, and lets you answer them in the app.
  • Command palette (Ctrl+K) with projects, terminals and shortcuts; search terminal content with Ctrl+Shift+F.
  • Project card with the branch and worktrees, with protected removal.
  • CLAUDE.md and AGENTS.md are written only with your consent, folder by folder, with a preview.
  • Images in the terminal: paste a screenshot with Ctrl+V or drag up to 16 images to the agent.
  • Project cards in the order you want: drag them by the top or use Move up and Move down in the ··· menu.
  • Rename a terminal by double-clicking the pane name; the name is saved with the pane.
  • Free Use License: free for individuals and companies, including commercial use.
Verify installer integrity

Run the command below in PowerShell and compare its result with the published SHA-256. The hash verifies the file; it does not replace a digital signature.

Get-FileHash .\PrumoGrid-Setup-0.5.11.exe -Algorithm SHA256
991cd4c85f86157e181e3c3b132c5e36d4aec08462716f581c02d3ff6fe7198e

Application license

Free use

Starting with 0.5.11, PrumoGrid is distributed under the PrumoGrid Free Use License. Individuals and companies may install and use the app free of charge, on as many computers as they like, including at work and for commercial purposes. You may pass on the original installer, unmodified and free of charge. Selling, modifying or reverse engineering it is not allowed.

Read the full license ↗Text published with the installer ↗Third-party notices ↗

04 / THE DETAILS

Before your first prompt.

Security and privacy ↗
Do I need to use an AI agent?

No. PrumoGrid also works as a terminal grid organized by project. Agents are optional and must be installed and authenticated on your computer.

What happens when I close the window?

In persistent mode, terminals keep running in the local service and reconnect when you reopen. Parts of step coordination need the window to stay open. Restarting Windows stops the processes; resuming a conversation depends on the CLI, account and available history.

Is the source code available?

Source code is not distributed with this release. This site offers only the installer, under the PrumoGrid Free Use License 1.0: free for individuals and companies, including commercial use.

Does the app run locally?

Yes. PrumoGrid organizes terminals and data on your computer without requiring its own account. Agents and commands you run may use the internet and send data to the services you choose.

How do I install a plugin?

Open Tools → Extensions, install a .prumoplugin file, review its information and permissions, enable it and run its commands. Plugins run code with your account privileges: only install plugins you trust. VS Code/VSIX extensions are not supported.

How do I start after installation?

Open PrumoGrid, click add projects and select a folder. Open a terminal and work as usual. To use a CLI, run its command. When you want to organize tasks, use New step.

What data does this website collect?

We measure how the site is used anonymously, without cookies and without identifying visitors: page views, downloads, videos watched, language and approximate region. IP addresses are not stored, and the numbers are kept only as totals, for up to 90 days. Google Analytics is only turned on if you allow it; you can change that choice below, and downloads work the same either way. Your language preference is saved only in your browser. The contact form emails your message to the author through the Resend email service; it is not stored on the site. To block bots, the form uses Cloudflare Turnstile, loaded only when you start filling it in.

CONTACT

Let’s talk about how you work.

Have a suggestion, found an issue or want to share how you use PrumoGrid? I’m Carlos Alvarenga, the project’s creator.

So I can reply.

Up to 4,000 characters.

Your message is emailed to the author and is not stored on the site.

Carlos Alvarenga

Creator of PrumoGrid

Connect on LinkedIn ↗carlosalvarenga.dev ↗