You have the system, the plugins, every note you ever took — and you still spend the first twenty minutes of every AI session explaining where things stand. Silicon Cortex keeps one continuously-updated, human-verified picture of your work in plain Markdown on your own machine, so Claude, ChatGPT, Codex and Cursor all start where you left off. Nothing resets. Everything compounds.
The Gumroad founding presale has ended. No purchases are available through this site.
Product vision · 59 seconds: user-owned project memory, human approval and traceable decisions. Illustrative scenario.
If you run several serious things at once — a company, a product, a creative practice — and you work with more than one AI, you already know the tax: every new chat starts from zero. You re-explain your project, your constraints, your history. Again.
Claude doesn't know what you told ChatGPT. Codex starts from nothing. The smarter the models get, the more this one bottleneck costs you. You've become your own context clipboard.
Organise life and work as long-running tracks — a business, an album, your health — instead of a folder graveyard.
One Python file, zero dependencies, bring your own key. Turns quick captures into structured, traceable memory, keeps a live dashboard of focus and open loops, and answers questions about your own history.
Ships with CLAUDE.md / AGENTS.md guides. If you already use Claude Code or Codex, your agent reads your vault natively and writes results back in.
Nothing enters your confirmed memory until you approve it. Every summary is auditable back to its sources. Your AI can propose; only you decide what's true.
Anything you tag sensitive is held back from cloud models by default. Plus a daily spending cap — a background assistant never surprises you with a bill.
A folder you own — readable in any editor, greppable, git-able, Obsidian-compatible. If we disappear tomorrow, your memory doesn't.
Immutable.What you captured is never rewritten by a machine.
Always a draft.The AI proposes structure, summaries, connections — nothing more.
A separate layer only you can write to.Slower by design. Still readable and trustworthy 12 months in.
That's the whole product.
The architecture is the subject of a UK patent application, filed June 2026.
A company, a music practice, and this product itself are operated on this exact system every day.
Cloud calls go only to your chosen provider with your key. The vault never leaves your machine. We run no servers.
Silicon Cortex remains the founder's focus. The earlier Lite presale, pricing tiers, seat limits, delivery dates and promotional offers are no longer available.
This site now provides background on the system and its principles. There is no active purchase offer or new delivery commitment on this page.
Silicon Cortex is a reliable cognition layer that lives on your machine, in plain Markdown, and gives every AI you use — Claude, ChatGPT, Codex, Cursor — the same continuously-updated, human-verified picture of your work. No app, no account, no lock-in.
No — if you can open a terminal and paste three commands, you're in. A full install guide is included, written for humans.
Your own API key. Typical daily use is cents, not dollars, and Keeper has a built-in daily spending cap.
OpenAI and Anthropic out of the box; the provider is a one-line config field. Or skip API keys entirely and use your existing Claude Code / Codex subscription as the engine.
The vault is plain files on your disk. Only what you choose to process is sent, only to your provider, and anything tagged sensitive is held back by default. Nothing touches our servers — we don't have servers.
macOS and Linux natively; Windows works via WSL.
No. The Gumroad founding presale has ended. Previous prices, delivery dates and promotional offers are no longer available.
Silicon Cortex remains focused on context that users own and control.
Contact the founder