A personal AI operating system. Clone it, run Claude Code, answer a few questions, start working. Claude remembers who you are across every session.
Every time you open an AI tool, you start from zero. You re-explain your role, your work, your preferences. Every. Single. Time.
ContextOS solves this. It gives Claude Code persistent knowledge of who you are, what you work on, and how you like to work — so you get useful output from the first prompt of every session.
- No more re-explaining yourself. Claude knows your role, your tools, your current projects, and how you prefer to work.
- A prompt workshop you own. Save your proven prompts in
my-prompts/. Experiment without polluting shared team space. - Diagnostic skills. When output misses the mark, ask Claude to "load prompt-framework" — it'll ask which of the five parts (identity / task / context / constraints / output format) is missing.
- Shared knowledge that stays in sync. Compliance frameworks, tool inventory, approved prompts, workflows — maintained once in a team repo, available to everyone via git submodule.
- Institutional knowledge that doesn't walk out the door. Workflows and prompt patterns outlast any one person.
- ~10 minutes to set up. Clone, run
claude, answer questions in a conversation. - Zero maintenance burden. Team context updates with one command. Personal context is just markdown you edit when things change.
┌─────────────────────────────────────────────────────────┐
│ Your Personal OS (this repo, forked per user) │
│ │
│ CLAUDE.md ← auto-loaded every session │
│ my-context.md ← your role, tools, goals │
│ skills/ ← reusable processes │
│ my-prompts/ ← your prompt workshop │
│ │
│ shared-context/ ← optional git submodule │
│ ┌─────────────────────────────────────────────────┐ │
│ │ Team Shared Context (separate repo) │ │
│ │ Team knowledge, workflows, approved prompts │ │
│ └─────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────┘
| Repo | What It Is | Who Updates It |
|---|---|---|
| contextOS-personal (this repo) | Template for your personal AI context | You (your fork) |
| contextOS-team | Team knowledge base, workflows, prompts | Team via PRs |
The team context is included as a git submodule. Claude Code reads it seamlessly — no special config needed. You can skip it if you're a solo user — ContextOS Personal works standalone.
git clone --recurse-submodules https://github.com/botz-pillar/contextOS-personal.git ~/context-os
cd ~/context-os
claudegit clone https://github.com/botz-pillar/contextOS-personal.git ~/context-os
cd ~/context-os
claudeA SessionStart hook detects you haven't onboarded yet. Claude greets you and walks through a 7-step setup:
| Step | What it covers | Time |
|---|---|---|
| 1 | Domain & Identity | 1 min |
| 2 | Role & responsibilities | 2 min |
| 3 | Tools & systems | 2 min |
| 4 | Current projects | 2 min |
| 5 | Skills & learning | 2 min |
| 6 | Working style | 1 min |
| 7 | File generation | — |
Claude writes your personalized CLAUDE.md and a detailed my-context.md. From then on, every session picks up where your context left off.
Alternative (non-interactive):
python3 setup.pyruns a terminal questionnaire. Note:setup.pyis legacy (ships v1 template shape); primary path is the SessionStart hook +skills/onboarding.md. Both produce working files, butsetup.pyoutput may diverge slightly from the new template shape. Use it only if you can't run the interactive flow.
After setup:
/verify # confirm your setup is wired correctly
/plan # plan a work session (Claude uses your projects + priorities)
/draft # draft an email, report, or document in your voice
/research # structured research brief with recommendations
/review # critique your work before sending
/whats-new # see what changed in shared team context
Or try a natural prompt:
- "Summarize my role and current priorities" — confirms Claude knows you
- "What workflows are available in shared-context?" — discover what the team has built
- "Help me prepare for my meeting with [stakeholder] about [topic]" — uses your project context
CLAUDE.md (short, routing-focused) + my-context.md (detailed depth) describe you. Claude Code auto-loads CLAUDE.md every session.
shared-context/ is a git submodule pointing to your team's repo. One git submodule update --remote pulls the latest.
skills/— loadable procedures (onboarding,prompt-framework,chunking)shared-context/workflows/— team-standard procedures for specific workshared-context/prompts/— role-specific prompt libraries
.claude/commands/— slash commands (/plan,/draft, etc.).claude/hooks/— automation (first-run detection, submodule health check)
~/context-os/
├── CLAUDE.md # Your personalized AI router
├── my-context.md # Your detailed context (role, tools, projects, goals)
├── my-prompts/ # Your prompt workshop
├── skills/ # Reusable processes
├── shared-context/ # Team knowledge (if joined to a team)
├── .claude/
│ ├── commands/ # Slash commands
│ ├── hooks/ # Automation
│ └── settings.json # Permissions + hook config
└── examples/ # Reference setups (domain-specific)
git submodule update --remoteWhen someone on the team PRs an update to the shared repo, every member gets it with that one command.
Before rolling out to the team, customize the team repo:
git clone https://github.com/botz-pillar/contextOS-team.git
cd contextOS-team
claudeClaude detects the uncustomized team context and walks you through setup — team structure, standards, tool inventory, guardrails. Commit and push. Team members get the customized context automatically.
- Clone:
git clone --recurse-submodules https://github.com/YOUR-ORG/contextOS-personal.git ~/context-os - Run:
cd ~/context-os && claude - Answer the 7 setup questions (~10 minutes)
- Review generated files — edit anything that's off
- Try a workflow: ask Claude to load
shared-context/workflows/[relevant].md - Read
docs/collaboration-guide.mdfor contributing back to the team
| Guide | Covers |
|---|---|
| Setup Guide | Detailed walkthrough |
| Customization Guide | Tuning your personal context |
| Collaboration Guide | Contributing to shared team context |
| Common Mistakes | Seven traps to avoid when setting up |
Reference setups in examples/. Currently:
examples/infosec/— SOC Analyst, Cloud Security Engineer, Compliance Manager
More domain flavors coming. Contributions welcome — see docs/collaboration-guide.md.
MIT — use it, fork it, make it yours.