Your Second Brain, Now with a Mind of its Own.
Cortex is the first AI-native knowledge OS. Frictionless voice capture, durable AI agents, and a self-organizing graph. Stop organizing your notes — let Cortex do it for you.
Capture. Connect. Compound.
Three moves, one flywheel. Everything you say becomes structure; structure becomes connections; connections become work your agents finish for you.
Capture
Say it once — voice, web clipper, or a forwarded email. It lands as clean, structured Markdown before you've pocketed your phone.
Connect
Cortex threads every new thought into your graph — entities, tags, and bi-directional links, stitched in automatically like seams of gold.
Compound
Agents work the graph while you're away — researching, drafting, auditing — and deliver to the channels you actually live in.
Speak. We’ll handle the rest.
Brain dumps shouldn’t be hard. Just hit the mic or use the web clipper. Our AI instantly transcribes, structures into Markdown, extracts action items, and tags your note automatically.
- Raw voice → structured Markdown in under 3 seconds
- Action items extracted and routed to your task queue
- Auto-tagged and linked into your graph, zero filing
Don’t just store knowledge.
Put it to work.
Cortex features an enterprise-grade AI Agent Builder powered by Vercel Eve. Create custom subagents that run 24/7 — on your notes, on your schedule.
Research Agent
Tell it to find context while you sleep. Wake up to sourced briefs linked straight into your graph.
Drafting Agent
Have it write your weekly newsletter from your messy notes. Your voice, your sources, zero blank page.
Multi-Channel
Agents notify you via WhatsApp or SMS when tasks are done. Your second brain reaches you anywhere.
One agent is an assistant.
A swarm is a staff.
Every Cortex agent can spawn subagents — scoped workers that fan out, do one job well, and report back. Your Chief of Staff plans the run, delegates, audits the results, and only pings you when a human call genuinely matters.
Durable runs
Runs survive restarts and flaky APIs. Failed steps retry with backoff; nothing silently dies.
Scoped permissions
Each subagent gets exactly the notes, tools, and budget its job needs — and nothing else.
Approval gates
Nothing outbound ships without your tap. Gates are default-on for email, posts, and payments.
Replayable traces
Every run keeps a full trace — every source read, every decision made. Audit it, replay it, fork it.
Start with a staff,
not a blank page.
Prebuilt agents tuned across thousands of runs. Install one, point it at your graph, and make it yours — every prompt, trigger, and channel is editable.
Weekly Review
Your week, closed out: wins, open loops, and what Monday-you needs to know first.
Meeting Debrief
Every call becomes actions, owners, and a two-line summary — inside 90 seconds.
Inbox Triage
Reads the noise, surfaces the three emails that actually matter, drafts the replies.
Market Watch
Overnight movement in your tickers and your niches — sourced, scored, and cited.
Content Repurposer
One good idea becomes a thread, a post, a script, and a newsletter section.
Reading Digest
Everything you clipped this week, distilled — with citations back to every source.
Connect the dots
you didn’t know existed.
With native bi-directional linking and vector semantic search, Cortex visually maps your thoughts. Ask your graph a question, and our local RAG pipeline gives you the exact answer.
A second brain that
keeps its mouth shut.
Local-first by architecture, not by promise. Your vault lives on your device, sync is end-to-end encrypted, and your keys mean your models.
Local-first vault
Notes live on your device in plain Markdown. Offline isn't a mode — it's the default.
End-to-end encrypted
Keys are generated on-device and never leave it. Not even we can read your graph.
Bring your own keys
Plug in OpenAI or Anthropic keys and pay wholesale. Zero markup on inference, ever.
Never trained on
Your data trains nothing. Contractually, permanently, at every tier — including Free.
Built for people whose
ideas outrun their filing.
I stopped taking notes and started talking. Cortex turned six months of voice memos into a system I actually use.
The Market Watch agent pays for the subscription by Tuesday. Everything after that is compound interest.
First tool where the graph isn't a party trick. I asked what I decided about pricing in March — it knew.
Start free. Scale your mind.
Local-first, privacy-first. Upgrade when your brain does.
For getting your thoughts out of your head.
- Local-first sync
- Core Editor
- 50 Voice Captures / mo
- Graph view
For minds that don’t stop at 5pm.
- Infinite Voice Capture
- Custom AI Agents
- WhatsApp Notifications
- Semantic search + local RAG
- Priority sync
Bring your own OpenAI or Anthropic keys.
- Everything in Pro
- Wholesale token pricing
- Model of your choice
- Zero markup on inference
Everything else
you're wondering.
Does Cortex work offline?
Fully. Your vault is local-first — capture, edit, and search work with zero connection. Sync and agent runs pick up automatically when you're back online.
Which models power the agents?
Frontier models by default, routed per task — fast models for triage, deep models for drafting and analysis. On BYOK, every agent runs on your own OpenAI or Anthropic keys at wholesale cost.
Can agents act without my approval?
Not outbound. Approval gates are default-on for anything that leaves your vault — emails, posts, messages. Internal work (research, drafts, linking) runs freely; you control the gates per agent.
What happens to my data?
It stays yours. Device-resident vault, end-to-end encrypted sync, no training on your content at any tier, and one-click export to plain Markdown whenever you want out.
Can I import from Notion, Obsidian, or Roam?
Yes — Markdown-native importers preserve your links and structure. Cortex re-threads everything into the graph on import, so your old vault arrives already connected.
Is there an API?
Pro includes webhooks and a REST API — pipe captures in from anywhere, trigger agent runs programmatically, and pull structured results back out.
Upgrade your intellect.
Join the early access list — the first 500 builders get Pro free for 3 months.