SecBrain v1.0 · macOS · Apple Silicon

A mind,
just for you.

A personal AI that quietly organizes your calendar, messages, notes, and health — entirely on your Mac. Open source. Yours to inspect.

Download for macOS secbrain/secbrain·2.4k ★

Local-first by defaultOpen source · MITTamper-evident audit logLLM Inference provider: zero retention

secbrain.app/dashboard

Tue · 26 May · 11:04

Good morning, Vinicius.

● Today's brief · 7h ago

Calendar is light today — room to clear what's open. Sandra wants help with gov.br and hinted at a visit. Amor's birthday is Wednesday, in 3 days.

Agents

1running

Syncing data · 1m 54s ago

Tasks today

↑ vs week (avg 4)

● Inbox · urgent

Sandra Mara · 1d open

Goals · active

4 on track · 3 attention

How it works

Local on your Mac. Cloud when you need it.

SecBrain runs entirely on your Mac. Pick local LLM inference for total privacy, or route hard prompts through SecBrain Cloud — our zero-retention service — with sensitive data redacted before egress and every byte recorded in your audit log.

Indexed locally

Calendar, mail, contacts, notes, files. Everything stays on disk in an encrypted SQLite database. We never touch your data over the network.

~/Library/Mail
→ encrypted index
SQLite · AES-256 · ~/Library/SecBrain/db

Redacted before egress

Names, contacts, amounts, phone numbers, addresses — replaced with placeholders before any prompt leaves your device. Tier 3 data (medical, financial) asks you first.

"Sandra Mara" → Person A
"+55 48 99201-1083" → [phone]
"$2,400" → [amount]

Zero-retention LLM inference

We are the LLM inference provider. No logs, no caching, no training on your prompts — written into our own terms, not borrowed from somebody else's SLA. Every call is recorded in your local audit log with the exact tokens that left your device.

● egress_decision · remote
● redacted · tier 3 → allow
● chain verified · sha256 a3f7…b921

What it does

A second mind that earns your trust.

Not a chatbot. Not a productivity dashboard. A continuously running personal AI that learns your life and quietly keeps it organized.

Daily brief

A morning summary in your language. People mentioned, money discussed, dates coming up, threads needing a reply.

Unified inbox

Threads, tasks, and habits collapsed into one ranked stream sorted by what matters most today.

Goal-aware

Pin a horizon ("schedule volleyball reservations"), pin a why ("regular sports + social"). SecBrain links every task to a goal.

Composable agents

Write your own — schedules, prompts, delivery to WhatsApp / email / push. Or use the 33 system agents that ship with the app.

15+ connectors

Calendar, Mail, Contacts, Files, WhatsApp, Notion, Linear, Jira, Strava, Oura — all indexed locally on first sync.

Auditable everything

Every prompt, every redaction, every egress event — cryptographically chained. Verify the chain anytime in Settings.

Security

Built so we can't see it even if we wanted to.

Privacy isn't a checkbox on a page. It's a chain of design decisions you can inspect, in source code we publish under MIT.

The path your data takes

On your Mac

Encrypted SQLite, indexed, summarized, ranked.

↓

Redaction layer

Names, contacts, amounts replaced before egress.

↓

SecBrain Cloud (Pro build only)

We are the provider. Zero retention. No logs. No training on your data.

↓

Audit log

Every step recorded, SHA-256-chained, locally verifiable.

What we guarantee

No telemetry by default. The app is silent until you explicitly opt in.
Tier 3 is redacted, then routed. Medical, financial, intimate — names, amounts, identifiers replaced before egress. Or switch the agent to local LLM inference to keep them on-device.
Open source. MIT-licensed. Audit the redactor, the firewall, the storage layer.
Contractual zero retention. SecBrain Cloud has no-logs, no-train, no-cache built into its terms.
Source available. The whole desktop app — redactor, firewall, agents — is on GitHub for you to audit.

Pricing

Simple, honest pricing.

Pay only for what you use. Local LLM inference is free. Cloud LLM inference is billed per token — $10 minimum top-up. Earn $5 in credits for every friend you refer.

Recommended

SecBrainlocal or cloud · pay-as-you-go

The SecBrain app. Runs LLM inference locally or in our zero-retention cloud — pick per agent, per prompt, or globally. New features land here first.

$0.50/ 1M tokens · in

$0.50 / 1M tokens out. No subscription. Minimum top-up $10. Set a monthly cap.

Local or cloud LLM inference. Run any local model for free, or route to SecBrain Cloud for frontier-grade reasoning when you need it.
Runs on any Mac. No 64 GB requirement — pick local for the prompts your machine can handle, cloud for the rest.
Redaction before egress. Names, contacts, amounts replaced — same engine as SecBrainOSS.
Zero-retention contract. We don't keep logs, we don't cache, we don't train.
Email support · response under 24 h.

Get started · $10 minimum How billing works

What does $10 buy? Roughly 10 million tokens — ~50 daily briefs a day for a month, or ~1,000 deep reasoning answers. Refer a friend to earn $5 in credits each. Set a hard monthly cap so you never overspend.

Building from source? SecBrainOSS is the source-available variant. Clone it, build it on your Mac, and run with local LLM inference. For developers and security researchers who want to audit every line.

View on GitHub

Refer a friend

$5 for you. $5 for them.

When someone signs up for SecBrain Cloud with your link and makes their first $10 top-up, you both get $5 in credits — roughly 5 million tokens each. No cap on referrals.

Get your referral linkOnce you have an account.

How it works

You share your unique link.
They sign up for SecBrain Cloud + make their first $10 top-up.
Both accounts get $5 in credits the moment that first top-up clears.
Credits drop into your account instantly. No cap, no expiry, no catch.

FAQ

Questions, honestly answered.

What is SecBrain vs SecBrainOSS?

SecBrain is the app you want. It runs LLM inference locally or in our zero-retention cloud — pick per agent or per prompt. Easy install, $5 free credit, new features land here first. SecBrainOSS is the source-available variant on GitHub for developers and researchers who want to clone, build, and run with local LLM inference only. Most people should use SecBrain.

How does billing work?

$0.50 per million tokens in, $0.50 per million tokens out — flat, no tiers. A typical daily brief consumes ~25k tokens (~$0.013). A complex multi-step reasoning answer ~100k tokens (~$0.05). You top up in $10 / $25 / $100 packs ($10 minimum). Earn $5 in credits per referral. Set a hard monthly cap in Settings so you never overspend. Unused credits roll over for 12 months. If you run out, the app pauses cloud calls until you top up — no auto-charge.

What gets redacted before going to the cloud?

People names, contact details (phone, email, address), monetary amounts, exact dates of birth, and government IDs — replaced with placeholders before any prompt leaves your device. Tier 3 content (medical, financial, intimate) is redacted with stricter rules and gets a confirmation prompt before each call, so you always know when it's about to leave. The full redaction logic is on GitHub — read secbrain/firewall/redactor.rs. If you want zero egress regardless of content, switch the agent to local LLM inference — or build SecBrainOSS yourself.

Who runs the LLM inference for SecBrain Cloud?

We do. SecBrain Cloud is our own LLM inference service, running on our own infrastructure with a no-logs, no-cache, no-train policy written into our terms. You don't pick the model — we curate which one runs and rotate as better ones ship. The audit log shows what left your device, the date, and the exact tokens. Our DPA is at secbrain.app/legal/dpa. If you don't want to use our cloud at all, run SecBrainOSS instead.

Is SecBrain available on iPhone / iPad?

Companion iOS app is on the roadmap. It will read from your Mac's encrypted index over end-to-end-encrypted sync — never directly from cloud connectors. Beta opens late 2026.

How is this different from Apple Intelligence or ChatGPT?

Apple Intelligence only sees data inside Apple's own apps. ChatGPT sees only what you paste into it. SecBrain runs continuously, indexes your whole digital life on disk, and gives every prompt full context — with you picking, prompt by prompt, whether it runs locally or in our zero-retention cloud.

What if I stop topping up?

There's nothing to cancel — SecBrain is pay-as-you-go. When your balance hits $0, cloud calls pause and the app silently falls back to local LLM inference. Nothing auto-charges. Your data stays on your Mac. Top up again any time.

Is SecBrain open source?

The SecBrain desktop app source is on GitHub for auditing — redactor, firewall, agents, connectors. The closed piece is our hosted LLM inference + billing service. SecBrainOSS is the source-available variant you can clone and build yourself; it omits any code that talks to SecBrain Cloud and is meant for developers who want to verify every line and run fully local.

A mind,just for you.