AI Corp - Experimental AI agent company dashboard

Experimental AI agent company dashboard for orchestrating workspace-scoped agents, tasks, roles, Telegram bots, cron jobs, and pipeline-style automation.

AI Corp is a living lab, not a polished product. Expect fast iteration, incomplete edges, and ideas that may shift as the project evolves.

What This Is

AI Corp explores what a modern “AI company” interface can look like when agents are treated as first-class team members.

The project combines:

• workspace-scoped AI agents with roles, skills, and personalities
• Kanban-style task management
• inter-agent messaging and approvals
• Telegram bot integration
• cron-based automation
• command execution inside a Docker sandbox
• pipeline and event views for experimentation around agent workflows

Why The README Framing Matters

This repo is intentionally presented as experimental because:

• the product direction is still being explored
• features may appear, change, or disappear quickly
• some flows are designed for internal iteration rather than polished onboarding
• the value is in the system design, not in pretending it is finished

If you want a concise mental model:

Workspace = a company
Agents    = team members
Tasks     = work items
Roles     = permissions + behavior
Telegram  = external interface
Crons     = recurring automation
Pipelines = structured agent workflows

Highlights

• Workspace creation and management
• AI agents with roles, skills, memory, and personality files
• Task board from Backlog to Done
• Direct agent-to-agent messaging
• Telegram bot support per agent
• Cron jobs for scheduled actions
• Role-based permissions for file, folder, and system access
• Command execution in a controlled Docker sandbox
• Pipeline and event views for more advanced orchestration experiments
• Workspace initialization from .aicorp.yml

Stack

• Frontend: React 19, React Router 7, Zustand 5, Tailwind CSS v4
• Backend: Node.js, Express 4, TypeScript 5.8, tsx
• Data: JSON persistence in ~/.aicorp/ and SQLite-backed pieces where needed
• UI: Radix UI, Lucide React, motion animations
• Visualization: @xyflow/react, D3.js
• Automation: node-cron, Telegram bots, sandboxed command execution

Quick Start

Prerequisites

• Node.js
• Docker Desktop or Docker Engine

Local Development

1. Install dependencies:

   npm install

2. Start the frontend:

   npm run dev

3. Start the backend:

   npm run dev:server

The app runs on:

• frontend: http://localhost:3001
• backend API: http://localhost:4000

Docker Workflow

Use make if you want a more guided workflow:

make bootstrap
make start
make restart
make stop
make status
make logs
make logs-web
make logs-backend

This setup is useful when you want:

• the backend on the host for workspace path access
• the UI in Docker
• background agent managers running with the server

Sandbox Command Execution

Agents can execute shell commands inside a Docker sandbox scoped to their workspace.

This experiment focuses on safe defaults:

• non-root container user
• workspace-only filesystem mount
• resource limits
• approval flow for risky commands
• optional network access control

Two permissions control this feature:

• system:run_commands
• system:approve_commands

Workspace Settings

Each workspace can define:

• whether command execution is enabled
• Docker image to use
• CPU, memory, PID, and timeout limits
• network access policy
• destructive command policy
• Git write policy

First-Time Setup

The first time you open a workspace:

• make sure Docker is running
• create or open a workspace with a valid folderPath
• let the system create the command sandbox automatically

If Docker is unavailable, command execution fails gracefully and the rest of the app still works.

Development

npm run dev
npm run dev:server
npm run lint
npm test

A Few Notes

• The project is optimized for experimentation and iteration, not production hardening.
• UX and data models may change as the experiment evolves.
• Some features are intentionally internal or opinionated to support rapid testing of agent workflows.

Project Structure

src/
  components/   UI and views
  lib/          shared client utilities
  server/       API, tools, automation, persistence
tests/          Vitest coverage for core logic