Now live: plug your notes into Claude Code & Cursor

A second brain that shows its work.

Upload docs, voice notes and images, then ask. NullBrain answers with a citation to the exact source — or says “that’s not in your notes” instead of inventing one.

Free forever tier · No credit card · One-tap Google sign-in

nullbrain — ask
What did I decide about pricing in the March planning memo?
You set Pro at a usage-based tier and agreed to revisit enterprise quotes after the pilot.

ChatGPT guesses. NullBrain quotes.

Lives in the tools you already think in
Claude Code Claude Desktop Cursor Telegram Obsidian Chrome Clipper YouTube MCP REST API Signed webhooks
Voice notes PDF & Word Highlight to save Voice clone Podcast digest Galaxy map Study mode Meeting copilot Team brains SSO / OIDC
Capabilities

Everything you save becomes something you can ask.

Not a chatbot bolted onto your files — a memory that cites itself, answers in your voice, and follows you into your editor, your chats and your meetings.

Answers with receipts

Every answer names the exact document, page or audio timestamp behind it. Tap the chip, jump to the source.

The client agreed to a one-week slip if voice ships complete — hard limit is the end of the month.
📎 call-recording.m4a · asr · @12:40

Your brain, inside your editor

One API key plugs your notes into Claude Code, Claude Desktop and Cursor via MCP.

$ claude mcp add nullbrain
✔ Connected
tools: search_notes · ask ·
add_note · get_digest

Honest when it doesn’t know

Below the relevance bar it stops and says so. It would rather refuse than invent a source.

“That’s not in your notes. I won’t guess.”

A copilot that sits in on meetings

It listens in, surfaces the right note at the right moment, and when someone asks what your brain already knows, hands you the answer with its source.

Capture from anywhere

Forward to Telegram. Highlight in Chrome. Paste a YouTube link. Import a Notion page. Drop a PDF. It all lands in one brain.

Obsidian, mirrored

Sync your vault with one command. Edit a note and your brain updates itself, no duplicates.

Study mode

Your notes quiz you back — spaced-repetition flashcards generated from anything you’ve saved.

Drafts that cite themselves

Turn your notes into an email, post or report — grounded in your brain, every claim carrying its source. You edit; it never invents.

Chat with one document

Lock the conversation to a single doc and ask away — answers stay inside it, cited line by line.

Agent tools, opt-in

Let it use web search, page fetch and a calculator on hard questions. Off by default, always cited.

Teams with real walls

Shared team brains with roles and SSO/OIDC, walled off by row-level security the database itself enforces.

One API, signed webhooks

Automate everything over REST. Events arrive HMAC-signed, so you can trust the sender.

The map

Your knowledge, rendered as a universe.

Hub concepts burn as suns; your notes orbit as planets. The map is yours alone and reloads exactly where you left it.

Your knowledge as a constellation A calm cosmic knowledge graph: note-nodes drift as stars connected by faint indigo constellation links. One bright focal node sends a dotted answer ribbon out to two emerald evidence receipt chips, showing its work. §1 §2

Live in the app — this one’s just a small system.

How it works

From brain dump to cited answer.

1.0

Capture

Drop PDFs, record voice notes, forward to Telegram, clip the web, sync Obsidian, paste a YouTube link, or push notes in over MCP and the REST API.

2.0

Constellate

Every chunk is embedded into your private galaxy and stamped with provenance — source, method and time you can audit.

source · method · timestamp
3.0

Ask

A cited answer in seconds — by text or voice. Or an honest “that’s not in your notes.”

Trust

Private by architecture, not by promise.

isolationRow-level security: tenant A can never query tenant B. Enforced in the database, tested both ways.
encryptionYour files are encrypted at rest with per-tenant keys.
computeSpeech, embeddings and your voice run on a private machine behind an authenticated tunnel — not a shared cloud GPU.
trainingYour notes never train anyone’s model. Period.
languageVietnamese-first: native VN and EN–VI code-switch, in text and speech.
0
API tests passing
0%
of answers cited — or refused
0
ways to capture a thought
Voice

It answers in your voice. Literally.

Record one short, consent-gated sample. Then any answer — and a daily podcast digest of your notes — can come back narrated by you. All of it on private hardware.

  • Consent-gated cloning — your voice, only by you
  • Daily podcast digest, read in your own voice
  • Full-duplex voice loop with barge-in

Try the voice loop →

“We built NullBrain because we didn’t trust our own AI notes. So we made every answer carry a receipt.”

The NullBrain team · building in public

Pricing

Start free. Upgrade when your galaxy grows.

Pay only when you want your own voice — or a brain for your team.

Free

Your whole second brain, free

$0/month
  • Cited answers and honest refusals
  • Docs, audio & YouTube capture
  • Telegram bot & web clipper
  • Galaxy map & study mode
  • MCP + API key
Start free
⭐ Most popular

Pro

Everything, plus your voice

Contact us
  • Everything in Free
  • Voice clone — answers in your voice
  • Daily podcast digest, narrated by you
  • Meeting copilot & agentic retrieval
  • Full-duplex voice loop
Go Pro

Team

A brain your whole team shares

Custom
  • Everything in Pro
  • Shared team brain, roles & invites
  • SSO / OIDC + SCIM
  • Priority support
Talk to us
FAQ

Fair questions, straight answers.

Can anyone else see my notes?

No. Every tenant is isolated with database-enforced row-level security, your files are encrypted at rest with per-tenant keys, and our test suite proves tenant A cannot read tenant B, in both directions.

What happens when the answer isn’t in my notes?

NullBrain tells you so. It can optionally answer from general knowledge, but that answer is clearly labeled and carries no citation. It never pretends your notes said something they didn’t.

Is my voice clone safe — and can I delete it?

Cloning is consent-gated and only works with your own sample. The sample is encrypted and processed on a private machine, never a shared cloud. You can delete it any time, and deleting your account removes everything.

Do my notes train anyone’s model?

No. Your content is used to answer your questions, full stop. Nothing you upload is used for training, ours or anyone else’s.

Can I export everything and leave?

Yes. Your originals stay intact, the API exposes your library, and deleting your account permanently removes your documents, embeddings, voice profile and history.

What does free actually include?

The full second-brain loop: capture from docs, audio, Telegram, Chrome and YouTube, cited answers with honest refusals, the galaxy map, study mode, and an MCP/API key. Voice cloning and the podcast digest are Pro.

Your notes are smarter than you think. Ask them.

Set up in two minutes: sign in, drop a file, ask your first question.

Want first access to what’s next — mobile apps and new experiments?

No spam. Invites and ship notes only.