Context windows are for sessions. Your codebase is a contract.

Iwo Szapar’s AI Second Brain gives your coding agent a repo of skills, hooks, adversarial reviewers, and SQL memory with decay scoring. Even a million tokens fills in one afternoon, and everything the session learned used to die with it. Now Monday's decision is injected into Friday's session start. Works with Claude Code and Codex.

See pricing

Inspectable code, not a black box · Your CLAUDE.md survives the install · One-time purchase

second-brain-dev · terminal or desktop, your call

LIVE

$ /recall "rate limiting"

→ searching indexed repos · 2 hits

api-service · Feb 12
Upstash Redis, sliding window, 100 req/min per IP

consulting-app
token bucket via next-rate-limit

api-service pattern fits your Vercel stack. Middleware adapted, applied to /api/login.
→ memory_knowledge +1

your own repos, indexed · nothing re-explained

The problem, in developers' words

“A goldfish with a PhD.”

A 1M window fills in one session, and your project runs for years. You know the workarounds because you maintain them:

session #47 · context1M tokens

⚠ compacted · everything learned: gone

SQL memory

survives every session · greps back

Bigger windows don't fix this. Sessions end. Memory shouldn't.

the actual problem

“I maintain a 500+ line CLAUDE.md that I manually update after every session.”

dev.to

“Spent three hours architecting the solution. Next morning: 'I don't have context from previous sessions.'”

claude-mem writeup

“It re-suggested the exact pattern I banned last week.”

context-rot thread

What you will actually use it for

Ship-speed jobs, memory included.

What /yalla, /recall, and the hooks actually do to a week.

Feature, end to end

/yalla reads your CLAUDE.md, plans phases, writes the migration and UI, and logs the decisions back to memory.

1 day → 2 h

per feature

beforeafter

Cross-repo recall

/recall answers from every repo's decision log: why Postgres, which auth pattern, what you banned and why.

re-explanations

beforeafter

Handoffs that write themselves

The decision log becomes the handoff doc: schema choices, gotchas, open threads. Generated, not remembered.

3 h → 15 min

per handoff

beforeafter

Token costs are real ($2-6/feature) · timings illustrative

What this actually is

Not another AI.
Memory for the one you have.

You keep your AI and your subscription. The Second Brain is the private memory it reads before answering, and writes back to after.

Works with the AI you already pay for

ChatGPT

OpenAI

Claude

Anthropic

Gemini

Google

Copilot

Microsoft

Pick any. Switch anytime: the memory is yours, not the vendor's.

01 · You ask

/recall auth decisions

in the AI app you already use

02 · It reads your memory

~/second-brain/ · plain files · your machine

decisions/auth.md

gotchas.md

conventions.md

+ what it learned → filed back

nothing leaves · nothing trains a model

03 · You get your answer

The why behind the code, the gotcha you hit in May, your conventions applied.

specific to your work, first try

Ask → read → answer → file back. Every session leaves the memory smarter.

plain files

Folders on your computer

Markdown you can open, edit, move, or delete without us. No database you can't see, no export button you have to trust.

skills + agents

Commands that know your work

Plain-English commands wired to your folders. No prompt engineering: the workflow is stored, you just call it.

write-back

It compounds

Decisions, promises, and preferences file themselves after each session. Month three answers better than day one.

Same prompt · Same model · Different memory

Vanilla agent vs. yours.

Three commands from the shipped repo. Illustrative transcripts, real mechanics.

Without memory

You: Add user profile page with avatar upload

I'll help you add a user profile page. Let me start by asking about your tech stack, existing auth setup, and preferred image hosting. Can you share your current database schema?

With your Second Brain

You: Add user profile page with avatar upload

Running /yalla... Plan: 4 phases (schema, upload endpoint, UI, tests). Read your CLAUDE.md: Next.js + Supabase + shadcn. Wrote migration, /api/avatar-upload with 5MB limit, ProfilePage.tsx. Tests: 6 Playwright specs pass. Review: 0 blockers. PR #442 opened.

Plan, build, test, review, ship. One command.

One tracked week · session replay

The week compounds. Watch it.

Scrub the days. The memory counter only moves one way.

memory rows: 1,206

$ /begin "avatar upload"

→ branch session/task-442 + issue #1122 created

$ /yalla "profile page with avatar upload"

plan: 4 phases · schema → endpoint → UI → tests

✓ 6/6 Playwright specs pass

✓ PR #442 opened, 0 blockers

→ memory_knowledge +2 (upload pattern, waitUntil usage)

1,206 → 1,210 in one week · no re-explaining on Monday

Who already bought

Six months of daily dogfood.

This website ships through this system. The commit history is public.

per /yalla feature · ~400K tokens

per /yalla-team feature · ~1.2M tokens

6 mo

of daily dogfood · commit history public

Runs on the AI plan you already pay for · no margin on your tokens

Wouter van den Bijgaart

AI Developer

Replaced DIY setup

“I built my own system and spent 3 hours a week maintaining it. The Second Brain does the maintenance automatically.”

Gabe Marusca

Developer · Founder

20+ hrs/week

“Now I type /yalla and ship features I used to spend a day on.”

Wytze

Engineer

DIY buyer

“That Puppeteer script is amazing. I repurposed it as a Proposal Generator.”

Tobias

Engineer · Power user

Runs workshops

“Been doing this local second brain thing with Claude Code for myself mostly. Have held professional workshops to teach others.”

Asked before buying

“I can build this myself.”

I can build this myself over a few weekends.

You can. This repo is what that build looks like after six months of daily iteration: the hooks that survived, the reviewer prompts that stopped hallucinating, the memory schema with decay scoring. Read the file tree before you decide.

Is this just a big CLAUDE.md?

No. CLAUDE.md is one file in the system. The repo ships skills (markdown), hooks (shell scripts that fire deterministically), agents (planner, implementer, reviewers), and a SQL memory backbone. Each layer is inspectable code you own.

Will it clash with my existing setup?

The questionnaire asks for your current CLAUDE.md, AGENTS.md, scripts, and hooks, and merges your rules into the generated repo. Your stack, review strictness, and commit cadence carry over.

Where does my code and data live?

On your machine. Default backend is PocketBase, local SQLite, offline-capable. Bring your own Supabase if you want sync. Your code, patterns, and review findings never leave your device.

Codex or Claude Code?

Both. The repo ships CLAUDE.md + AGENTS.md + .agents/skills + .codex/config.toml. Your client stays familiar. The shared brain gets sharper.

Forward deployed?

There is a dedicated page for engineers who rotate through client deployments: per-client memory, handoff docs generated from the decision log. Find it under Solutions as Second Brain for FDEs.

I already purchased Second Brain. What's new in 2.0 and do I need to upgrade?

2.0 is a structural rebuild. The AI agent reads your questionnaire and configures everything for you (no manual setup). The system now prunes stale knowledge automatically so it sharpens over time, and health monitoring flags when things drift before you notice. Your existing setup is a valid foundation and transfers directly. Email hello@iwoszapar.com to discuss upgrade options.

I'm not technical. Can I use this?

54% of our clients had never opened a terminal before. 58% are in non-technical roles. The setup is guided step-by-step, and the Kickstart option means our AI configures everything from your questionnaire answers. You don't need to write code.

How is this different from ChatGPT Memory or Projects?

ChatGPT Memory stores random facts. This engineers your entire work context — voice patterns, client history, decision frameworks, connected tools. It's the difference between AI that 'remembers your name' and AI that 'knows your business and runs your operations.'

What if AI models change? Will my system break?

Your context, skills, and workflows live in files and a database you control. Codex, Claude Code, ChatGPT Work, Claude, Gemini, and local agents can use that core through their own setup. Client-specific hooks and plugins need their matching adapter.

What do I need to get started?

A computer and an AI surface you want to use. Claude Code, Codex, and local agents run the full workspace setup. Chat and desktop surfaces use the connections and workflows available to them. The guided process configures the right starting point.

What's the difference between DIY, Kickstart, and Done-With-You?

DIY ($197): Full repo, self-serve setup, no support. MemoryOS available separately. Kickstart ($597): AI agent configures your system from a questionnaire, plus async support and first year of MemoryOS Pro. Done-With-You ($2,497): 2-hour onboarding call, full infrastructure deployment, and monthly updates.

Is there a money-back guarantee?

Yes. 7-day unconditional money-back guarantee on all packages. If it doesn't work for you, email us and we refund — no questions asked.

Wherever you are

Your existing tools. No extra layer.

CLAUDE.md + AGENTS.md + .codex/config.toml in one repo. Terminal or desktop app, Claude Code or Codex: same brain.

Your AI, as you already use it

Claude Code and Codex, desktop app or CLI, and on your phone through the apps already on it. Nothing new to learn, no new tab to live in.

No new monthly bill

It runs on the AI plan you already pay for. The system itself is a one-time purchase. Nobody wants subscription #14.

No lock-in, ever

Plain files and a local database on your machine. Stop paying nothing, lose nothing: it keeps working forever, and you can walk away with everything.

Battle-tested

100+ professionals run it in daily work, most of them non-technical. Six months of the creator’s own client work ships through it.

Point-tool competitors are another login, another subscription, another silo. This is neither another, nor either.

After you buy

Made yours, while you sleep.

Nobody gets a template. The questionnaire feeds a setup agent that researches your world and configures a system that is yours alone.

15 minutes of questions

Your clients, your tools, your writing samples, the shape of your week. Plain English, no technical questions.

An agent researches and builds yours

Configured uniquely for you on the current frontier models. No two repos come out alike. It runs overnight: Iwo does not need to be awake for your brain to get built.

Delivered in 24–48 hours

Folders named after your real clients, your voice rules, your commands. You open it and it already knows you.

Working in ~30 minutes

Your first session is a real briefing, not setup. 48-hour async support and a 7-day money-back guarantee stand behind it.

Included with Accelerator · Starter self-installs with the same questionnaire · Launch Partner adds the live 2-hour call

Trust

He has trained 25,000+ professionals at Microsoft, Walmart, Deloitte, PwC, and beyond.

Six months of daily iteration and thousands of commits. This website ships through the same pipeline, and the commit history is public.

Creator of MemoryOS · Co-created the AI Maturity Index with Harvard researchers

Iwo Szapar · The Builder

Pricing

Pay once. Own it forever.

Quick math

100+ professionals are running their Second Brain right now. If it saves you 5 hours per week at $200/hr, that's $4,000/month in recovered time. The Kickstart pays for itself in the first week. One purchase. No subscriptions.

Devs start here

For technical users who want full control

DIY

The full system. You set it up.

$197.00one-time

Full repo: 55 skills, 21 agents, 7 integrations

Setup guide + Health Check MCP

Community access

Self-serve — you configure at your pace

MemoryOS available separately ($199/yr)

No support included

7-day money-back guarantee.

Wytze

Engineer · DIY customer

That Puppeteer script is amazing. I repurposed it as a Proposal Generator.

Get DIY - $197.00

For professionals who want it working today

Kickstart

AI configures it. You're running in minutes.

$597.00one-time

Everything in DIY

AI agent builds your system from your answers

48-hour priority async support

First year of MemoryOS Pro included ($349 value)

Async support — no calls needed

7-day money-back guarantee.

Wouter

AI developer

I built my own system and spent 3 hours a week maintaining it. This does the maintenance automatically.

Get Kickstart - $597.00

For leaders who want it built for them

Done-With-You

2-hour onboarding call + full setup.

$2,497.00one-time

Everything in Kickstart

2-hour onboarding call with Iwo

I deploy your infrastructure (Vercel + Supabase)

Monthly Second Brain updates (new agents, skills, tools)

12 months MemoryOS included

2-hour call + direct access to Iwo

7-day money-back guarantee.

Damian Nomura

Done-With-You customer

This personal session was one of the best investments for me and my business.

Get DWY - $2,497.00

Not sure whether Done-With-You is the right fit? Take the fit check.

Your system stays current. Automatically.

New tools and skills ship through MemoryOS. Updates arrive without you doing anything. Health monitoring spots what needs attention. Weekly recommendations tell you what to fix. Included in Kickstart and Done-With-You.

Learn about MemoryOS

What Happens After You Buy

Get Repo

5 min

Instant email with GitHub repo

Questionnaire

15 min

Answer about role and workflows

Build

~60 min

System generates, configured for you

First Win

7 min

Run /overview, priorities are there

Deploy Infra

30 min

One-click Vercel + Supabase (optional)

Second Brain 2.0 · Developers

Install this afternoon.

The file tree is above. Read it, then decide. Your CLAUDE.md survives the install, and Monday's decisions stop dying on Friday.

See pricing

Your code never leaves your machine

One-time · from

$197

See pricing

Context windows are for sessions. Your codebase is a contract.

“A goldfish with a PhD.”

Ship-speed jobs, memory included.

Not another AI.Memory for the one you have.

Vanilla agent vs. yours.

The week compounds. Watch it.

Six months of daily dogfood.

“I can build this myself.”

Your existing tools. No extra layer.

Made yours, while you sleep.

He has trained 25,000+ professionals at Microsoft, Walmart, Deloitte, PwC, and beyond.

Pricing

DIY

Kickstart

Done-With-You

Your system stays current. Automatically.

What Happens After You Buy

Install this afternoon.

Not another AI.
Memory for the one you have.