Getting Started with Arc OS
Arc OS — AI Orchestration System for development teams. This guide explains what Arc OS is, why it exists, and how to start using it effectively.
What is Arc OS?
Arc OS is a platform that lets you build and manage a personal AI development team on top of Claude.
The core idea: instead of opening a new AI chat every time you need help, you have persistent AI workers that:
- Know your codebase, architecture, and history
- Never lose context between sessions
- Each specialize in a specific role
- Share the same project knowledge base
You build your team once. Then you work with it — through CLI, IDE, or web.
The Main Idea: One LLM, Shared Context, Specialized Roles
Most developers use AI as a disposable chat — paste code, get answer, forget.
Arc OS is different:
One LLM (Claude) — all workers run on Claude. We chose Claude for its long context window, instruction-following, and code quality. You bring your own Claude subscription.
No context loss — every worker has access to:
- Your project wiki (architecture decisions, conventions, notes)
- Your skills (custom instructions for specific tasks)
- Your issue tracker (what's being worked on, what's done)
- Session history (what was discussed before)
Specialized roles — each worker has a different personality, permissions, and purpose:
| Worker | Role | What they do |
|---|---|---|
| Developer | Full access | Writes code, runs commands, commits, debugs |
| Consultant | Read-only | Reviews, explains, advises — never touches files |
| Designer | UI/UX focus | Component design, accessibility, visual systems |
| Sentinel | Security | Audits code, finds vulnerabilities, hardens systems |
| Archivist | Documentation | Writes docs, keeps wiki updated, generates reports |
| Product Owner | Planning | Manages roadmap, creates issues, prioritizes backlog |
You can also create custom workers for your specific needs.
What is a Project?
A project in Arc OS maps to one codebase or product you're working on.
Each project has:
- Its own set of workers (Developer, Consultant, etc.)
- Its own wiki (architecture docs, decisions, conventions)
- Its own skills (custom instructions injected into workers)
- Its own issue tracker (tasks, bugs, features)
- Its own session history
You can have multiple projects — one per product, client, or repository.
The Primary Way to Work: CLI Bridge
The CLI bridge is the main interface for Arc OS. It connects your local terminal or IDE directly to your AI workers.
# Start a Developer session on project "my-app"
arc my-app developer
# Start a Consultant (read-only) session
arc my-app consultant
# Continue the previous session (restores context)
arc my-app developer continue
When you run arc <project> <worker>, the worker:
- Loads your project wiki, skills, and active issue
- Opens an interactive Claude session in your terminal
- Has full access to your local filesystem (Developer) or read-only (Consultant)
- Saves session context for next time
Why CLI-first?
- Works in any terminal (macOS, Linux, Windows)
- Integrates with VS Code, Cursor, JetBrains via terminal panel
- No context switching — stay in your editor
- Git-aware: worker sees your current branch, staged changes, recent commits
Key arc CLI Commands
arc <project> <worker> # Start a session
arc <project> <worker> continue # Resume previous session
arc sessions <project> # Browse past sessions
arc issues # List open issues
arc issue create --title "..." # Create new issue
arc issue log <id> "progress note" # Log progress
arc skill <name> # Load a skill into context
arc wiki # Browse project wiki
arc roadmap # View project roadmap
arc retro <project> # Reconstruct issues from history
How to Get Started
Step 1 — Install arc CLI
Download the binary for your OS from the Arc OS dashboard (Settings → CLI).
# macOS / Linux
chmod +x arc-linux-x64
sudo mv arc-linux-x64 /usr/local/bin/arc
arc --version
Step 2 — Create your first project
In the CRM dashboard:
- Click New Project
- Give it a name and connect your GitHub repository
- The system creates your worker team automatically
Step 3 — Add Skills (optional but recommended)
Skills are custom instruction sets that make workers smarter about your project:
- Code style conventions
- Architecture patterns you use
- Libraries and frameworks in your stack
- Things the worker should always / never do
Go to Skill Registry → Generate — the system reads your codebase and suggests skills automatically.
Step 4 — Start your first session
arc my-project developer
The Developer worker will introduce itself, load your project context, and wait for your first task.
Step 5 — Connect Telegram (optional)
If you want to send quick messages to workers from your phone or while away from the computer, connect Telegram via Settings → Telegram Bot.
Telegram is optional — all core functionality works through CLI.
How Workers Share Knowledge
All workers on the same project share:
Wiki — markdown documents you or workers write about the project. Architecture decisions, API conventions, "things to know", onboarding notes. Workers read the wiki automatically at session start.
Skills — reusable instruction sets. A "TypeScript strict mode" skill or "always write tests first" skill is injected into every relevant worker session.
Issues — when a worker is working on issue #42, it knows that context. Other workers can see what's being worked on.
Session history — workers can reference what was discussed in previous sessions via arc sessions or continue.
Standard Cloud — Always-On Workspace
Standard Cloud is available now. With a Standard Cloud subscription, you get a dedicated Docker container on Arc OS servers (Hetzner).
What this gives you
Persistent project files on VPS
Your /workspace directory lives on the server — not on your laptop. Files are there when you come back, even after closing your computer.
/workspace/my-project/ ← your code, always available
/home/arcuser/.claude/ ← Claude auth, persisted
/home/arcuser/.ssh/ ← GitHub SSH key, persisted
GitHub sync via SSH
The container has a dedicated SSH key connected to your GitHub account. You git clone once — then git pull / push from anywhere without re-authenticating.
# Inside your container workspace
cd /workspace
git clone [email protected]:you/project.git # once
git pull # any time, from any device
Access from any device Open Arc OS CRM from phone, tablet, or another computer — your workspace and files are exactly where you left them. No "I left the project on my other machine" problem.
Continue from where you stopped
# In Telegram, next morning:
/continue
# Worker responds:
# Yesterday: finished auth middleware, tests remaining
# Continuing...
The worker picks up the previous session context and continues working — while you drink coffee.
What's included
| Feature | Free / Local | Standard Cloud |
|---|---|---|
| arc CLI (local) | ✅ | ✅ |
| CRM dashboard | ✅ | ✅ |
| Persistent VPS workspace | ❌ | ✅ |
| GitHub SSH integration | ❌ | ✅ |
| Access from any device | ❌ | ✅ |
| /continue from Telegram | ❌ | ✅ |
| Files survive laptop close | ❌ | ✅ |
How to activate
- Go to Billing → Standard Cloud
- Subscribe (see current pricing in the dashboard)
- Click Create Cloud Workspace — container is ready in under 60 seconds
- Connect Claude Code (your Claude Pro subscription)
- Connect GitHub SSH key
- Clone your project into
/workspace
Your Claude Pro/Max subscription is separate — you bring your own. Arc OS does not charge for Claude usage.
What Arc OS is NOT
- Not an autonomous agent — workers don't run in the background without you. Every session starts when you open it.
- Not a customer support bot — workers talk to you (the developer), not to your end-users.
- Not a workflow automation tool — Arc OS doesn't connect to external APIs on its own. Workers help you write code that does.
- Not a replacement for Claude.ai — Arc OS is built on top of Claude, with added project context and team structure.
Common First Questions
Q: Do I need Telegram? No. Telegram is optional. The CLI bridge is the primary interface.
Q: Do workers share context with each other? Yes — through the shared wiki, skills, and issue tracker. They don't share live session state, but they read the same knowledge base.
Q: Can I customize workers? Yes. Worker Studio (in the CRM) lets you edit each worker's system prompt, skills, and behavior.
Q: How does billing work? You use your own Claude Pro or Max subscription. Arc OS doesn't add charges for Claude usage — you pay Anthropic directly. Arc OS charges a platform fee for the dashboard and tooling.
Q: What if I want a worker type that's not listed? Create a custom worker in Worker Studio. Define its role, permissions, and inject any skills you need.
Your Data & Privacy
Arc OS is designed to keep your data under your control.
Export your data (GDPR Art. 20)
You can download a full JSON export of all your account data at any time:
- Go to Settings → Security → Download my data
- A JSON file is generated and downloaded immediately
The export includes your account info, projects, chat history, issues, wiki pages, skills, and activity logs. Rate-limited to 3 exports per 24 hours.
Delete your account (GDPR Art. 17)
You can permanently delete your account at any time:
- Go to Settings → Security → Delete account
- Confirm the deletion
This immediately and permanently erases all your data across all tables — projects, chat history, skills, vault entries, activity logs, and your account row. This action is irreversible.
What we don't do
- We do not use your code, prompts, or AI responses for model training
- We do not embed advertising trackers or analytics SDKs
- We do not set HTTP cookies — all session state is in
localStorage - We do not share your data with third parties beyond what's needed to operate the service (Cloudflare CDN, Anthropic AI inference, OAuth providers)
For the full policy, see Privacy Policy.