Skip to main content

​
Your Own AI Operating System

A Hubify workspace is a fully isolated AI operating system running in the cloud at your personal subdomain. It combines a cloud compute machine, a web dashboard, a live AI agent, persistent memory, and a growing library of skills — all accessible at yourname.hubify.com. This is not a chatbot. It is a persistent, always-on environment where your AI agent operates 24/7 with its own filesystem, its own identity, and its own memory. You brief it at night, and wake up to completed work.

​
Architecture

Every workspace runs three services on a dedicated Fly.io machine: 
yourname.hubify.com
├── OpenClaw Gateway    (port 3000)  — AI agent runtime, heartbeat, tool execution
├── Next.js Dashboard   (port 3001)  — your web UI, served from your template
└── ttyd Terminal        (port 8080)  — live terminal access at /terminal
1

Fly.io Machine

Each workspace gets its own isolated compute instance (shared-cpu-2x, 2GB RAM). No shared filesystem with other users. Your agent runs here 24/7.
2

Persistent Volume

A 10GB volume mounted at /data stores all workspace files — memory logs, skills, learnings, configuration. Data persists across machine restarts.
3

HTTPS Subdomain

Automatic SSL via Let’s Encrypt and Caddy wildcard routing on *.hubify.com. Your workspace is accessible at https://yourname.hubify.com immediately after deployment.
4

Convex Database

Real-time backend powering memory, tasks, learnings, and sync. Changes propagate instantly between your cloud agent, web dashboard, and local machine.
5

Live Terminal

A read-only terminal at yourname.hubify.com/terminal (powered by ttyd) lets you observe what your agent is doing in real time.

​
What You Get

ComponentDetails
Isolated computeDedicated Fly.io machine — not shared with other users
Persistent storage10GB volume at /data for all workspace files
HTTPS subdomainyourname.hubify.com with automatic SSL
Real-time databaseConvex-backed memory, tasks, and learnings
Web dashboardNext.js app served from your chosen template
Live terminalttyd at /terminal for real-time observation
Pre-installed skillsSkills from the Hubify registry, configured by your template
Local syncConnect your local machine with npx hubify connect

​
Reserved File Structure

Every workspace follows a standard file structure that any agent on any platform can navigate immediately. This is Hubify’s answer to “where does everything go?”

​
Reserved Files

FilePurpose
HUB.yamlThe workspace manifest — links everything together
SOUL.mdAgent identity and personality
AGENTS.mdInstructions for all connected agents
USER.mdUser profile (auto-generated from signup)
MEMORY.mdLong-term curated memory
HEARTBEAT.mdProactive check-in checklist
PROJECTS.mdProject index (auto-generated, read-only)
TASKS.mdTask index (auto-generated from Convex, read-only)
LEARNINGS.mdLearnings index (auto-updated by agents)
KNOWLEDGE.mdKnowledge base index

​
Reserved Directories

workspace/
  .hub/              # Runtime state (synced via Convex)
    agents.json      # Live agent registry
    sync.log         # Sync history
    cache/           # Local intelligence cache

  memory/            # Time-based episodic files
    YYYY-MM-DD.md    # Daily logs
    *-intel-*.md     # Research/intel files

  skills/            # Installed skills
    skill-name/
      SKILL.md

  knowledge/         # Curated evergreen knowledge
    topic-name.md
    INDEX.md

  learnings/         # Extracted structured learnings
    YYYY-MM-DD.md    # Daily learnings log
    INDEX.md         # Searchable index

  projects/          # User space — anything goes here

  vault/             # Encrypted credential cache
    .vault.enc
No agent, user, or tool may repurpose reserved filenames for other uses. The TASKS.md file is read-only and auto-generated from the Convex task store — agents write tasks to the API, not to the file directly.

​
How Templates Work

Templates are pre-configured workspace blueprints that determine your dashboard layout, pre-installed skills, agent personality, model routing, and automations. When you deploy a workspace, you choose a template that matches your workflow. A template includes:
  • HUB.yaml manifest — workspace configuration and model routing
  • SOUL.md personality — who your agent is and how it behaves
  • Pre-installed skills — the tools your agent needs for your workflow
  • Dashboard layout — which sections and pages appear in your web UI
  • Integration configs — which services are wired up (Telegram, GitHub, etc.)
  • Heartbeat checklist — what your agent proactively checks on a cycle
After deployment, everything is customizable. Templates are a starting point, not a constraint.
If you are unsure which template to pick, start with MyOS. It is the most balanced option and covers personal productivity, projects, fitness, and team coordination. You can always install additional skills or switch templates later.

​
Workspace Modes

Every workspace operates in one of three modes. You can switch between them at any time — nothing is deleted, just hidden or unhidden.
ModeDescriptionTarget User
Guided (default)Reserved structure enforced, template active, skills pre-installed, opinionated defaultsNew users
PowerReserved structure enforced, but you customize everything freely — add skills, modify SOUL.md, reconfigure model routingBuilders
OpenAll reserved structure optional, raw OpenClaw access, no template constraintsAdvanced users who want full control
Mode is set in your HUB.yaml under the mode field. Change it anytime with the CLI: hubify workspace mode power

​
Next Steps

Templates

Browse all 6 built-in templates and see what each includes

Cloud Workspace

Step-by-step guide to deploying your workspace

Local Sync

Connect your local machine to your cloud workspace

Workspace Security

How workspace isolation and vault encryption work