Loci — Dive In

Your AI doesn't remember you.

LOCI

gives your AI a brain.

Open Source Local Markdown Works with Claude Code

"Restarted session — it remembered every decision from last week."

— @JasonZ, engineer

"Other AI memory is chaos. Loci is organized by design."

— @MikeL, full-stack dev

"Context full? Save, restart, back in 10 seconds."

— @YukiT, indie hacker

"It told me I started 3 side projects and finished none. Brutal but needed."

— @DevPatel, founder

"CLAUDE.md is a sticky note. Loci is a second brain."

— @SarahK, backend engineer

"Default memory is a messy notebook. Loci is a memory palace."

— @LiWei, AI researcher

"After Loci, my AI isn't a chatbot — it's my external brain."

— @RuiChen, designer

$ curl -fsSL https://raw.githubusercontent.com/codesstar/loci/main/install.sh | bash

★ Star on GitHub

The Problem

The endless loop.

Every session. Every project. Every context reset.
Your AI can write code — but it can't tell you
what to do today.

What if it didn't forget?

What if it woke up knowing your goals,
your decisions, your patterns —
and told you exactly where to start?

The Solution

Give your AI a real brain.

loci/ ├── me/ identity, values, preferences ├── tasks/ focus + queue + archive ├── decisions/ timestamped choices + reasoning ├── journal/ daily, weekly, monthly summaries ├── planning/ daily plans + calendar ├── links/ cross-project connections ├── dashboard/ visual command center ├── plan.md life direction & goals ├── inbox.md quick capture └── CLAUDE.md AI behavior rules Plain Markdown. On your machine. Git-versioned. Yours forever.

Designed for agents, not humans browsing folders.
The AI reads and writes. You just talk.

How It Works

Three layers. Load only what matters.

┌───────────────────────────────────────────────────┐ │ Layer 1 — Always Loaded │ │ CLAUDE.md plan.md inbox.md auto-memory │ │ Your AI knows who you are from the first token. │ ├───────────────────────────────────────────────────┤ │ Layer 2 — On Demand │ │ Tasks, daily plans, people, project docs. │ │ Loaded when you enter a topic. Context stays lean.│ ├───────────────────────────────────────────────────┤ │ Layer 3 — Deep Storage │ │ Archive, old decisions, evolution history. │ │ Never deleted. Surfaces when you need it. │ └───────────────────────────────────────────────────┘

◇

It remembers.

Decisions, preferences, context — saved as structured Markdown. Distilled, never dumped.

◦

It plans your day.

Morning briefings. Pattern detection. "You estimated 2 days — your last 3 took 2x that."

⇄

It connects projects.

Mistakes in Project A become warnings in Project B. Decisions propagate everywhere.

↑

It grows with you.

Day 1: knows your name. Month 3: your patterns. Year 1: how you've changed.

In Practice

Without Loci $ claude > I'm Callum, a data science... [new session] > Remember, I prefer Python... [context full] > *sigh* Starting over.

With Loci $ claude [loading plan.md, active.md, memory] > Morning, Callum. 2 focus tasks today. Start with the integration? [pattern detected] > You've started 3 side projects this month. Ship this one first?

Optional visual dashboard. Zero dependencies.

Today

Memory

Brain

Your AI is no longer a stranger.

Your goals. Your patterns. Your growth.
Across every project. Every reset.
And all of it belongs to you.

$ curl -fsSL https://raw.githubusercontent.com/codesstar/loci/main/install.sh | bash

★ Star on GitHub

Works with Claude Code · Codex CLI · OpenClaw · Cursor

Read the Docs →