ADG-Parallels is a VS Code extension concept that turns your editor into a
distributed AI workforce with hierarchy, automation, self-healing, and
a hint of corporate nonsense.

Think of it as:

• Kubernetes, but for Copilot.
• A tiny startup inside your VS Code windows.
• A company where you are the CEO and all employees are polite little AI agents
  (internally known as EggBots™, formerly “Ejajkas”).

It coordinates multiple AI sessions, delegates work, splits mega-tasks, and watches over workers like a very overcaffeinated project manager.

This repository currently contains the vision, roadmap, and corporate statute of this… organization.

Yes, we wrote a statute for AI employees. And yes, they follow it.
We live in the future.

Because Copilot (and every other LLM assistant) can:

• solve one task at a time,
• in one window,
• with one brain cell.

…while you are paying for 8+ parallel sessions that just sit there doing nothing.

We fix that.

ADG-Parallels aims to:

• spawn multiple VS Code windows (“workers”)
• delegate tasks between them
• auto-start and auto-continue Copilot
• detect when they finish or hang
• restart zombie workers
• and allow a Manager AI to coordinate everything

Basically:
Copilot Swarm → but make it corporate and cute.

Inside ADG-Parallels, every VS Code window becomes a character in our little corporate sitcom:

Workers behave.
Managers gossip.
Team Leaders pretend they know what's going on.
CEO wonders why everything is taking so long.

Corporate realism at its finest.

CEO (You)
│
▼
Manager AI (Claude Opus)
│
├── Team Leader AI(s)
│ │
│ └── Worker 1…N (VS Code windows)
│
└── Worker Pool (GPT-4o, 4o-mini, etc.)

All communication happens through a shared folder:
.adg-parallels/ — the secret underground HQ of all EggBots™.

• Automatic project provisioning
• Spawn N workers in separate VS Code windows
• Auto-start Copilot with start prompts
• Auto-resume work when Copilot finishes
• JSON task queue with statuses
• “No tasks left → close worker window gracefully”

• Dashboard for monitoring workers
• Pause/Resume entire workforce
• Adapter system (task plugins)
• Mega-task splitting (“write 100 articles” → auto-generated subtasks)
• Heartbeat & self-healing (detecting frozen workers)
• Hierarchical delegation (Team Leaders managing sub-workers)

Full details in:

• PROJECT_VISION.md
• PROJECT_ROADMAP.md
• CORPORATE_STATUTE.md (yes, we have a statute) 🥚

Each task has a type, e.g.:

• article-generation
• translation
• code-audit
• task-splitter
• generic

Adapters define:

• start prompts
• output rules
• completion criteria
• how to save files
• retries & validation

Workers simply follow orders like good little digital employees.

The CEO can say:

“Write 100 cooking articles.”

And the Manager AI will:

• detect mega-task
• split into 100 subtasks
• assign them to workers
• collect results
• merge everything back

This is basically project management, except nobody complains about meetings.

Every worker sends a small “I’m alive” ping every 30 seconds.

If a worker doesn’t ping:

• Manager marks it as unresponsive
• Closes the window
• Reassigns the task
• Launches a fresh worker

A fully automated “IT guy walking around restarting computers”.

.adg-parallels/
management/
project_tasks.json
hierarchy-config.json
attachments/

worker/
worker-config.json
worker-start-prompt.md
worker-continue-prompt.md
.heartbeat.json

adapters/
article-generation.adapter.json
translation.adapter.json
code-audit.adapter.json
task-splitter.adapter.json

jobs/
worker_1/
worker_2/
worker_3/

Workers are literally launched into their own tiny universe (folder).
It’s adorable.

• AI power users
• Copilot Pro users who want more than “one tab at a time”
• Content creators
• Agencies requiring mass automation
• Developers doing parallel refactoring / audits
• People who enjoy yelling “GO LITTLE WORKERS, GO!” at their monitor

🟡 Concept Phase — Implementation Begins Soon
This repository currently contains documentation, architecture and early planning.

Once the scaffolding is ready, the real fun begins (worker orchestration, dashboards, adapters, hierarchy…).

Contributions, ideas, and feedback are welcome!
Especially from humans (EggBots™ are not allowed to contribute to the repo yet — HR said no).

If you find this concept interesting, want to collaborate, or just want to say hi:

• Open an issue to discuss ideas
• Star the repo to show support ⭐
• Watch for updates — implementation is coming soon!

We're building this fast, but good ideas are always welcome.

TBD — likely MIT or AGPL.

EggBots™ get very excited when they see stars.
(It’s in their corporate statute. Probably.)