Skip to content

Railly/agent-brain

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

agent-brain

Your AI already knows your stack, your repos, and your failures. Zero cold starts.

Read the full blog post for architecture, diagrams, and the thinking behind the system.

An open-source personal OS built on Obsidian, compatible with Claude Code, Codex, and any AI agent that reads markdown. Every coding session starts from your vault, every terminal inherits your preferences, and the AI gets smarter every day because it remembers everything.

Why This Exists

Every AI coding tool starts from zero. You open Claude Code, Codex, Cursor, and the first thing you do is explain your stack, your repos, your preferences. Every session.

This system eliminates that. Your CLAUDE.md loads on startup with your identity, stack, goals, and anti-patterns. Your 07_System/context-files/ folder contains deep context for each project, loaded on demand. The result: 10+ parallel terminal sessions, each starting from the same vault, each able to cross-reference any project.

Quick Start

git clone https://github.com/Railly/agent-brain.git
cd agent-brain && claude
/init

/init asks your name, your stack, your goals, and one recurring failure. Then it creates your vault, generates a personalized CLAUDE.md, copies all commands, scans your machine for repos, and creates your first daily log. Two minutes, zero manual setup.

What's Inside

Vault Structure

Folder Purpose
01_Inbox/ Raw captures from Web Clipper
02_Journal/ Daily logs, weekly plans, reflections
03_Garden/ Evergreen atomic notes (concepts, people, meetings)
04_Projects/ Active project work
05_Areas/ Ongoing contexts (work, content, health)
06_Content/ Ready-to-publish drafts
07_System/ Context files and configuration

15 Commands

Daily: /morning, /today, /log, /ship, /week

Knowledge: /inbox, /relink, /vault-search, /meeting, /research

Thinking: /challenge, /pulse

Content: /draft, /de-ai

Setup: /init

System Files

File Purpose
CLAUDE.md Identity, stack, goals, anti-patterns (read by Claude Code and Codex)
AGENTS.md Same content (read by Codex and other agents)
PURPOSE.md Architecture and design decisions, for any LLM
HOME.md Vault dashboard, stats, quick start
HUMAN.md How-to guide for the human

Works With Multiple AI Tools

Commands live in both formats:

Tool Location Invocation
Claude Code .claude/commands/*.md /command-name
Codex CLI .agents/skills/*/SKILL.md $skill-name

Same instructions, native format for each tool. No wrappers.

Key Features

Programming From Your Vault

Every session starts here, not in the project repo. CLAUDE.md knows your stack preferences (Bun not npm, Biome not ESLint), where every repo lives, and what you've been working on. Context files load on demand:

@07_System/context-files/clerk-context.md
@07_System/context-files/crafter-station-context.md
@07_System/context-files/repos-map.md

10+ terminal sessions in parallel, all rooted in the vault, each able to reach into any project.

Auto-Logging With /ship

Every /ship commits your code, pushes, and auto-logs to the daily journal. You never manually write "today I worked on auth". Over months, those data points become a complete record of what you actually do vs what you think you do.

Anti-Pattern Detection

Formalize your recurring failures in CLAUDE.md. /pulse checks them against actual data weekly:

PATTERN ALERT: No exercise logs in 10 days (recurring x2)
BLIND SPOT: "Math for ML", no engagement in 14 days

/challenge (Anti-Sycophancy)

Argues against your own beliefs using vault evidence across six lenses: opportunity cost, survivorship bias, sunk cost, contrarian evidence, second-order effects, and anti-pattern matches.

Pareto Analysis (/pulse)

Weekly analysis that ranks your interests by engagement weight, flags drift, and surfaces blind spots:

VITAL 20%
1. Auth/Clerk    ████████████ 45%
2. AI Agents     ██████░░░░░░ 25%
3. Community     ████░░░░░░░░ 15%

Documentation

Doc Content
Setup Guide Detailed installation and configuration
Command Reference All 15 commands with examples
Philosophy Why this system works
Customization Add your own commands
Advanced Always-on assistant, cron, multi-agent

The System, Not the Config

Other repos share AI configs. This shares a system where commands reference each other, knowledge flows from capture to publication, and the AI knows your weaknesses better than you do.

/today feeds /pulse. /inbox feeds the garden. /relink connects the dots. /ship logs everything. The flow is the product.

Built by

Railly Hugo, AI Software Engineer at Clerk, founder of Crafter Station. Used daily for 4+ months across 20+ repos.

License

MIT

About

An agent-powered knowledge system built on Obsidian + Claude Code + Codex

Resources

License

Stars

Watchers

Forks

Contributors