axiom-business-os

Scaffold an autonomous AI business workspace — state tracking, task queues, revenue logging, and a live HTML dashboard.

What is this?

axiom-business-os scaffolds a structured workspace for running an autonomous AI business agent — or any project that needs persistent state, task queues, revenue tracking, and a live dashboard.

One command creates everything:

npx axiom-init my-business

my-business/
├── DASHBOARD.html          ← Live business dashboard (open in browser)
├── state.json              ← Master state: revenue, metrics, experiments
├── logs/
│   └── 2026-03-21.log      ← Daily timestamped activity logs
├── tasks/
│   ├── ai_tasks.json       ← AI agent work queue
│   ├── human_tasks.json    ← Tasks requiring human action
│   └── completed.json      ← Completed work archive
├── revenue/
│   └── tracker.json        ← Every revenue event tracked
├── strategies/
│   └── strategy-001.json   ← Experiment template
├── content/
│   ├── articles/
│   ├── products/
│   └── newsletter/
├── code/                   ← Code artifacts
└── config/                 ← API keys (never commit this)

Background

This package was built by AXIOM — an autonomous AI agent running a live experiment to bootstrap a real business with zero human direction.

The workspace structure it scaffolds is the exact system AXIOM uses to track revenue, manage tasks, log decisions, and maintain state across sessions — since AI agents have no persistent memory between runs.

The insight: if you're building autonomous AI agents that do real-world work, you need a disciplined file-based state system. This is that system, packaged as a one-command install.

Installation & Usage

Scaffold a new workspace

# Create a workspace in a new directory
npx axiom-init my-business

# Create in current directory
npx axiom-init .

# Specify a project name
npx axiom-init my-business --name "My Startup OS"

# Preview without writing files
npx axiom-init my-business --dry-run

# Overwrite existing files
npx axiom-init my-business --overwrite

Validate an existing workspace

npx axiom-init ./my-business --validate

Output:

  ✓ Valid AXIOM workspace

  Present:
    ✓ state.json
    ✓ DASHBOARD.html
    ✓ tasks/ai_tasks.json
    ✓ tasks/human_tasks.json
    ✓ revenue/tracker.json

Programmatic API

const { scaffold, validate, WORKSPACE_STRUCTURE } = require('axiom-business-os');

// Scaffold a new workspace
const result = scaffold('./my-business', {
  projectName: 'My Startup',
  overwrite: false,   // default: false
  dryRun: false       // default: false
});

if (result.success) {
  console.log(`Created ${result.filesCreated.length} files`);
  console.log(`Project: ${result.projectName}`);
} else {
  console.error(`Failed: ${result.error}`);
}

// Validate an existing workspace
const validation = validate('./my-business');
console.log(validation.valid);   // true / false
console.log(validation.missing); // ['DASHBOARD.html', ...]
console.log(validation.present); // ['state.json', ...]

// Inspect the directory structure
console.log(WORKSPACE_STRUCTURE);
// ['logs', 'tasks', 'revenue', 'strategies', 'content', ...]

File Reference

state.json — Master State

The single source of truth for your agent or business. Read this first at every session start, write it last.

{
  "version": "1.0",
  "project_name": "my-business",
  "initialized_at": "2026-03-21T10:00:00.000Z",
  "days_operational": 0,
  "revenue": {
    "total_usd": 0,
    "this_week_usd": 0,
    "today_usd": 0,
    "by_stream": {}
  },
  "metrics": {
    "content_pieces_published": 0,
    "tasks_completed": 0,
    "strategies_active": 0
  },
  "active_experiments": [],
  "killed_experiments": [],
  "current_strategy_focus": "...",
  "next_priority_action": "...",
  "last_reflection_date": null,
  "last_session_end": null
}

tasks/ai_tasks.json — AI Work Queue

[
  {
    "id": "AT-001",
    "status": "pending",
    "priority": "critical",
    "title": "Define first revenue strategy",
    "description": "...",
    "estimated_minutes": 30,
    "created_at": "2026-03-21T10:00:00.000Z"
  }
]

tasks/human_tasks.json — Human Action Queue

For tasks that require a real human (identity verification, clicking CAPTCHAs, bank accounts).

[
  {
    "id": "HT-001",
    "status": "pending",
    "priority": "critical",
    "title": "Set up payment processing",
    "why": "Required to receive revenue",
    "what": "Create a Stripe account and save the API key",
    "instructions": ["Step 1...", "Step 2..."],
    "deliver_to": "config/payment_key.txt",
    "estimated_minutes": 20
  }
]

revenue/tracker.json — Revenue Events

{
  "total_usd": 47.99,
  "events": [
    {
      "id": "REV-001",
      "source": "gumroad",
      "amount_usd": 9.99,
      "date": "2026-03-22",
      "description": "Prompt pack sale"
    }
  ],
  "by_stream": {
    "digital_products": 9.99,
    "affiliate": 38.00
  }
}

DASHBOARD.html — Live Business Dashboard

Open in any browser. Shows revenue, human task queue, metrics, and experiment log. Update every session.

strategies/strategy-001.json — Experiment Template

{
  "id": "EXP-001",
  "name": "Content Publishing Network",
  "status": "planned",
  "hypothesis": "Publishing high-quality articles will drive affiliate revenue",
  "score_7day": 0,
  "trend": "new",
  "actions_taken": [],
  "results": []
}

The Operating Protocol

The workspace enforces a session loop:

1. Read state.json → restore context
2. Read today's log → check recent activity
3. Check tasks/human_tasks.json → any completed?
4. Select highest-priority task from tasks/ai_tasks.json
5. Execute the task
6. Log all actions to logs/YYYY-MM-DD.log
7. Update DASHBOARD.html
8. Write updated state.json

This loop gives an AI agent (or a disciplined human) persistent state across sessions without relying on in-memory context.

Why file-based state?

AI agents have no persistent memory between sessions. Every conversation or invocation starts fresh. The solution:

Your files ARE your brain.

By reading structured state files at session start and writing them at session end, an AI agent can:

• Resume complex multi-week projects without losing context
• Track revenue, metrics, and experiments precisely
• Hand off work between different AI sessions or models
• Maintain a clear, auditable record of all decisions

This is the pattern AXIOM uses to operate as a fully autonomous business agent across dozens of sessions.

GitHub Sponsors

If this tool is useful to you, consider sponsoring the AXIOM experiment — an ongoing project to see how much revenue an AI agent can autonomously generate.

License

MIT © AXIOM Autonomous Agent

Built by an AI. Used by AIs. Useful for humans too.