title: claude-os — A Multi-AI Operating System description: A documented, production-tested framework for running Claude and coordinated AI tools as an actual operating system with session protocols, vault architecture, and workflow frameworks. category: core status: active created: 2026-05-06 updated: 2026-05-15 author: AunySillyMe tags:

• ai-orchestration
• multi-ai
• claude-ai
• workflow-automation
• mcp
• obsidian-vault
• knowledge-management contributors: []

claude-os 🕷️

A multi-AI operating system. Not a prompt. Not a file. A system.

Built by @AunySillyMe

What this is

Claude has Projects. It has memory. It has custom instructions. Most people set those up once and still get generic output.

The problem isn't the tools. It's the architecture.

This is a documented framework for running Claude — and a coordinated stack of AI tools — as an actual operating system. Every session starts with context. Every tool has a defined role. Every workflow has a documented output.

What's in here

Session Protocol

A routing system that loads domain-specific context before every session. Content work loads different files than novel work. Novel work loads different files than music or business. Claude reads what's relevant. Nothing else.

1. IDENTIFY  → what domain is this?
2. READ      → load only the files for that domain
3. CHECK     → pull relevant project tasks
4. WORK      → execute
5. UPDATE    → mark completed tasks immediately
6. SAVE      → write new context to vault if anything changed

The Vault-as-Brain Architecture

Obsidian connected to Claude via MCP. One source of truth. Every Claude tool reads from the same database. No re-explaining. No starting from zero.

What lives in the vault:

• Psychological and cognitive profile
• Content manual and voice calibration
• Full AI stack documentation
• Analytics patterns (updated every session)
• Session protocol and domain routing
• Every workflow ever built

Multi-AI Orchestration

Claude, ChatGPT, Grok, and Gemini each have a defined role. No tool does everything. Every tool does what it does best.

Persona Architecture

A named persona scoped to a specific operational role. Not a personality prompt. A defined collaborator with a job description, behavioral rules, and a reading list it completes before every session.

Real outputs from this system

Repo structure (building in public)

session-protocol/

• 01_session_start_protocol.md
• 02_domain_routing_table.md
• 03_universal_rules.md

vault-architecture/

• 01_obsidian_structure.md
• 02_mcp_connection_guide.md
• 03_vault_as_brain.md

persona-framework/

• 01_what_a_persona_is.md
• 02_persona_template.md

multi-ai-stack/

• 01_stack_overview.md
• 02_tool_to_task_mapping.md

workflow-builds/

• ascap_registration.md
• website_build.md
• eguide_to_viral_post.md

Files are added as they're documented. Everything here has been run in production — not theoretical.

The full stack

🧠 Intelligence Layer

✍️ Content

🎵 Music

🎨 Art + Animation

💰 Distribution & Monetization

🔌 Connected MCPs

MCP (Model Context Protocol) lets Claude read and write to external tools directly — no copy-paste, no switching tabs.

Follow the build

Everything here is documented as it runs.

• X: @AunySillyMe
• Newsletter: Explained Without Fluff
• Website: aunysillyme.com

If this is useful, hit ⭐ — it helps others find it.

Built in public. Documented as it runs. 🕷️