AI-OS Lite
An AI Personal Operating System — a Claude Code plugin that turns an Obsidian vault into an AI-operated daily workflow: morning briefings, voice-journal processing, evening reflection, task orchestration, and project status — all driven by your own notes.
AI-OS Lite reads and writes your vault, syncs work items to Todoist and Linear, and keeps your daily/project hubs current. It's the engine behind a frontmatter-routed, AI-navigable knowledge system.
Built by RandomStateLabs. AI-OS Lite is the open, public edition of the AI-OS system.
What it does
Seven skills, invoked by natural language or slash commands:
| Skill | What it does |
|---|
start-day | Morning prep — writes a "Yesterday" recap + today's context to your journal before you record |
process-journal | Processes a morning journal entry — extracts summary, mood, priorities, people; pushes priorities to Todoist |
prep-evening | Evening prep — writes the day's accomplishments and a wind-down prompt |
process-evening | Processes an evening entry — mood, habits, gratitude, tomorrow's tasks, pattern detection |
project-sync | Refreshes a project hub's status from recent devlogs, notes, and Linear |
vault-commit | Groups uncommitted vault changes into clean semantic git commits |
vault-config | Shared operational config — tool patterns, integrations, search strategy (not user-invoked) |
Requirements
- Claude Code
- An Obsidian vault following the AI-OS frontmatter conventions (see
AGENTS.md in your vault — the plugin reads your vault's structure, taxonomy, and project identity from there)
- Optional integrations: Todoist CLI (
td), Linear MCP, Obsidian local REST API
Install
# Add this marketplace
claude plugin marketplace add RandomStateLabs/ai-os-lite
# Install the plugin
claude plugin install ai-os-lite@ai-os-lite-marketplace
Then configure your vault path and conventions in your vault's AGENTS.md. The plugin substitutes your real areas, projects, contacts, and work-item prefixes from there at runtime — the skill examples ship with generic {WorkArea} / {Project} / {Contact} placeholders that resolve to your data.
Configuration
AI-OS Lite is per-user generic by design. It contains no hardcoded names, employers, or projects — every example uses a {Placeholder} token that the model fills from your AGENTS.md at output time:
{WorkArea} / {SideArea} → your areas{Project} → your projects (read from your vault filesystem){Contact} → people in your 4. Contacts/People/{TICKET} → your Linear team prefix
If a placeholder can't be resolved (no work area, no projects, no Linear team), the relevant example is simply omitted. Nothing breaks.
Licensing
AI-OS Lite is licensed under the GNU AGPL v3.0.
In plain terms:
- ✅ Free to use for personal and internal use — install it, run it on your own vault, modify it for yourself, no obligations.
- ✅ Free to study and modify — the source is open.
- ⚠️ If you modify AI-OS Lite and offer the modified version to others as a network service, AGPL requires you to make your modified source available to those users under the same license. (Normal personal use never triggers this.)
- ❌ You may not take this code, fold it into a closed/proprietary product, and distribute or host it for others without complying with the AGPL — or obtaining a separate commercial license.
Commercial licensing
Want to use AI-OS Lite in a way the AGPL doesn't permit — embedding it in a proprietary product, offering it as a hosted commercial service without open-sourcing your changes, or any other commercial arrangement? A commercial license is available.
Contact RandomStateLabs → github.com/RandomStateLabs to discuss terms.
Copyright © 2026 Yandi Farinango / RandomStateLabs. "RandomStateLabs", "AI-OS", and "AI-OS Lite" are names of the project authors; AGPL §7 does not grant trademark rights to use them to endorse or promote derived works.
Contributing
Contributions are welcome. Because AI-OS Lite is offered under AGPL and a separate commercial license, contributors will be asked to agree to a Contributor License Agreement (CLA) granting RandomStateLabs the right to license contributions under both — this is what keeps the dual-licensing model legally sound. (CLA process to be published alongside the first external contribution.)
