I stopped relying on scattered chats to remember my projects.

I built Personal Agentic OS on top of Obsidian so Codex, Claude Code, and Gemini CLI can work from the same local project context. Its dashboard shows project status and LLM usage using data from my own files.

The public version contains no project data. After installation, it reads only local sources you explicitly approve.

01Projects and decisions
02Local memory
03Three agents, one context
CodexClaude CodeGemini CLI

Three real views from my system

What used to be scattered now works as one system.

These are not product mockups. They are three working parts of the system I use for my own projects.

01Real system view
A real Obsidian memory graph
02Personal details removed
A real dashboard with sensitive values removed
03Real system view
A real preview, apply, and rollback view
01 / 03Keep memory in Obsidian
01

Keep memory in Obsidian

Approved project notes, decisions, and session summaries stay in your local Obsidian Vault instead of being locked inside one model’s chat history.

02

See evidence-backed status in one place

The dashboard updates project status, LLM usage, and budgets from approved local sources. Missing evidence stays Unavailable.

03

Review changes before applying them

Recommendations appear as a preview first. Applying a change and rolling it back each require separate approval.

01Real system view
A real Obsidian memory graph

Keep memory in Obsidian

Approved project notes, decisions, and session summaries stay in your local Obsidian Vault instead of being locked inside one model’s chat history.

02Personal details removed
A real dashboard with sensitive values removed

See evidence-backed status in one place

The dashboard updates project status, LLM usage, and budgets from approved local sources. Missing evidence stays Unavailable.

03Real system view
A real preview, apply, and rollback view

Review changes before applying them

Recommendations appear as a preview first. Applying a change and rolling it back each require separate approval.

One guided prompt, from system check to your local dashboard

  1. 01
    Inspect

    The agent checks Git, Node.js, and Obsidian.

  2. 02
    Choose

    You choose the Obsidian Vault, LLM tools, project folders, and permissions the installer may use.

  3. 03
    Approve

    Before anything changes, you review the files, skills, dependencies, and destination paths.

  4. 04
    Run

    The first sync fills the dashboard only with data found in approved local sources.

Run the setup prompt in any of these coding agents

CodexAGENTS.md
Claude CodeCLAUDE.md
Gemini CLIGEMINI.md

Paste the setup prompt into your coding agent

The prompt works with Codex, Claude Code, or Gemini CLI. It checks your system first and asks before installing software, writing files, or enabling telemetry.

master-install.md
Memory  → Obsidian Vault
Agents  → Codex / Claude / Gemini
Data    → Approved local files
Unknown → Unavailable
Changes → Review before apply
Download setup prompt

Before setup

01Does Personal Agentic OS send my project data anywhere?

The default setup keeps project names, file paths, Vault content, prompts, conversations, usage data, costs, and credentials on your device. Optional anonymous telemetry is off by default and requires your consent.

02What does this website track?

This website records aggregate page views, GitHub clicks, prompt copies, and downloads, along with language and campaign parameters. It does not use cookies or persistent user identifiers.

03What if Obsidian is not installed?

The setup agent first explains why Obsidian is used, then asks before installing anything. Dashboard-only mode is available, but shared project memory will not be.

04Can Codex, Claude Code, and Gemini CLI use the same project memory?

Yes. When you enable their adapters, each tool reads approved project context from the same local Obsidian Vault.

05Does the dashboard contain sample numbers?

No. Missing evidence remains Empty or Unavailable instead of becoming demo data.