Private beta · waitlist open →

Your intellect, current.
On your own hard drive.

A second brain that's actually yours — built from the AI you already use, kept current on your own device by Apple Intelligence, and working every day to further your passions and show you where your life can grow.

chanio captures every session you run — Claude, ChatGPT, Gemini, local. It compiles them nightly to plain markdown on your own disk, and loads the result into whichever AI you open next. Memory grows. Context shrinks. Cost falls.

See a sample brief — 30 sec, no signupJoin the private beta

No account. No API key. chanio works with the AI you already pay for — or free on your iPhone with Apple Intelligence. The key is only for power users who want chanio to run the model itself.

Already have AI chat history? Backfill it →Bring a decade of your own writing →
ClaudeChatGPTGeminiyour disk · nothing leaves6:00 AM · ON YOUR DEVICEA new angle on your passionscloud · bypassed

Your AI sessions → your own disk → compiled overnight into one graph → you wake up with a new angle. Nothing leaves your device.

▸ SEE IT RUN — IN THREE SECONDS

Paste any AI conversation. Watch chanio write the capture.

This is what chanio writes to your disk every session — the arc, the decisions, the artifacts, the names, the open loops, the durable IP. Multiply by 30 days and that's what your morning brief is reading tomorrow. No key required; capped at 8 KB — and it runs entirely in your browser, as you type. Nothing you paste here is sent anywhere.

Your passions — one per line
The capture below is weighted toward these. This is the file that aims your AI — the same ~/.chanio-focus chanio reads every night. It stays on your disk; nothing about it is stored here.
Your conversation
1,257 / 8,000 chars
The chanio capture
Paste or type on the left — the capture writes itself here as you do. The sample is a real cooperative-economics exchange; replace it with any conversation or article of your own.

What you just saw is one session. chanio runs this on every session you have with any AI, then a nightly loop compiles them into one morning brief on your disk. The capture is yours; the brief is yours; the AI you used is your choice.

▸ STOP RENTING AI. START OWNING IT.

You're paying $40–100 a month for AI memory. Just to the wrong people.

Claude Pro $20 · ChatGPT Plus $20 · Gemini Advanced $20 · Notion AI $20 · Cursor $20. Every subscription rents intelligence on every keystroke, locks your history inside their walls, and resets your context the moment a new model ships. chanio is the other architecture: routes by judgment, not by subscription.

Local · $0
Capture, extraction, canonical synthesis — runs on your hardware. The routine work that used to live in your subscription bill.
Cloud · ~$0.50/day
Morning brief synthesis only — the one step where frontier judgment compounds. Bring your own key. Cap your spend.

Your memory stays on your disk in plain markdown — queryable forever, portable to any AI that comes next. And the economics invert: because your memory is distilled locally, each AI call needs a small slice instead of your whole history. Memory grows. Context shrinks. Cost falls. Every conversation makes the next one cheaper — and better aimed, because the AI sees your distilled passions instead of your noise. The moat isn't the model. The moat is what compounds.

▸ See the $0.50/day routing →▸ Pricing — Free + Pro $9.99/mo →
▸ ALLOCATE YOUR AI SPEND

Your hard drive is a cost-allocation instrument. Pay for judgment. Store memory for free.

Subscriptions charge you flat — for memory and intelligence bundled, whether you use them or not, forgotten the moment a new model ships. chanio moves the memory to your disk at $0 and meters the intelligence to only the frontier calls you actually make. Toggle what you rent now, then set your real frontier usage.

What you rent now
Frontier-judgment calls you make
The calls you route to a top model. Capture, recall, and nightly synthesis of everything else stay on your disk at $0.
1/day10/day40/day
Renting now — flat$60/mo
Same charge whether you use the memory or not.
Allocated with chanio — metered~$9/mo
Memory: $0, on your disk. Judgment: 10 calls/day × ~$0.03. Scales with use, not with rent.
You reallocate $612/year away from flat memory-rent — toward the judgment you actually use, and toward nothing at all for the context that now lives on your own hard drive.

Illustrative. Assumes ~$0.03 of frontier judgment per call you route to the cloud (the site's own measured range is $0.01–0.05); capture, recall, and nightly synthesis run on your hardware at $0. The founder runs a full 25-project stack this way for ~$1.40/day. See the routing that makes it real →

▸ THE LEVERAGE LAYER

chanio is yours. The apps you make from it are yours. The agents you run on it are yours.

The health system that emerges around it answers to you. Keep your intellect current and recent from the AI of your choice, with as little cloud as necessary, so a local model becomes enough for what your life actually needs.

Your apps
Your accumulated intellect is the substrate. Ask the AI of your choice to build tools that fit your life — not the median life. Apps as utilities, not subscriptions.
Try the lab →
Your agents
Sovereign agents — they live where you live, hold your identity, call services on your behalf, bring the results home. Not corporate agents extracting your data on someone else's behalf.
See the overnight loop →
Your health system
You stop working for the system. Your chanio holds the truth of your history; your agents do the navigation; care comes home to where you actually live.
Visit co-op.care →
Apple Intelligence capable
On iOS 27 and macOS Tahoe, chanio routes through Apple's on-device model and Private Cloud Compute — local first, cloud cryptographically private when needed, your intellect stays in your trust boundary.
▸ SEE YOURSELF IN IT

Paste what you've written. Watch it reflect you back.

A journal entry, loose notes, a paragraph you've carried for years. chanio surfaces what you keep returning to, the question you keep circling, a thread you set down — and the convictions your own words already assert. It reflects only what you wrote; it doesn't guess your politics or profile you. Runs entirely in your browser — nothing you paste here is sent anywhere.

Write a few lines above and your portrait appears.

This is the shape — computed heuristically, in your browser, from one paste. The real chanio runs a model over your whole corpus: your books, your journals, ten years of clippings. This is a paragraph. Imagine a decade →

▸ THE OVERNIGHT — ON YOUR OWN DEVICE

You sleep. Your passions wake up smarter.

Everything you pulled from your AI today lands on your hard drive. Overnight, agents work it against what you care about — finding the patterns you didn't name, connecting sessions you had days apart, turning scattered curiosity into a brief of new angles. You wake up, and your passions have moved forward without you.

Your morning brief6:00 AM · computed while you slept · on your device
A pattern you didn't name
You circled the same tension three times this week across two different chats — here it is stated plainly, with the one question that unlocks it.
A connection across sessions
What you asked one model on Monday and another on Thursday are the same problem. Overnight it built the bridge you didn't have time to.
An edge to chase
One source, one idea, one next move — picked because it advances what you actually care about, not what's trending.
Runs on Apple IntelligenceThe overnight work runs on your iPhone and Mac's on-device model — with Apple's Private Cloud Compute only when a step genuinely needs more, cryptographically, no human in the loop. About $0 a night. Nothing leaves your device.

The cloud giants now offer to amplify and protect your intelligence — housed inside their systems. chanio takes the opposite vow: your intelligence never leaves your hard drive. There is nothing to protect it from, because we never hold it. The agents come to your memory. Your memory never goes to them.

Why sleep-time synthesis worksSee the on-device routing

A log of every curiosity you've ever had.

Every link you drop into Claude — a repo, a paper, a competitor, a half-formed question — is a curiosity. chanio logs them all, dates them, and threads the ones that belong together. Your past curiosities become a map; the next link you drop extends it.

JUN 29
journals.sagepub.com — Drane et al., Am J Health Promot 2026
50% of sandwich caregivers report serious suicidal ideation in the prior month. the cost of going unseen is not abstract. what does it mean when identification itself is the clinical intervention?
filed tohealth data infrastructure
JUN 28
ornith.ai — self-scaffolding RL, 35B MoE, MIT
a model that rewrites its own extraction prompt per task. is a personal knowledge graph better served by a model that adapts, or one that stays consistent?
filed toAI knowledge infrastructure
JUN 26
hhs.gov/onc — TEFCA crosses 1 billion health records
10M → 1B records in under a year. the interoperability pipe is real. but HIPAA stops at the hospital door — who owns the last mile of the data?
filed tohealth data infrastructure
JUN 26
newyorker.com — the AI design aesthetic taking over the internet
Claude Design defaults to the same palette for every user. does identity diverge from the tools, or do the tools flatten it?
filed toAI knowledge infrastructure
JUN 15
axios.com — loops as the next level of AI use
vital info → agents scout → I judge → repository improves. did VandeHei just describe chanio?
filed tohow chanio works
· · ·
↳ the next link you drop
chanio already knows which thread it joins.

A narrow capture — on purpose.

chanio only learns from what you choose to type into an AI. Nothing else. That's the honest line — and the reason the graph stays clean enough to trust.

What chanio sees
  • Prompts you send to Claude, ChatGPT, Gemini, Perplexity
  • The follow-ups and corrections that refine your thinking
  • The attachments you paste in — only with your consent
What chanio does not see
  • Your general browsing history or cookies
  • Your email, DMs, or social feeds
  • Anything you haven't explicitly sent to an AI
What a personal context graph actually does

Scattered sessions become
a queryable map of your thinking.

Every AI session you’ve ever had was useful in the moment. None of them talked to each other. chanio is the layer that connects them — turning isolated prompts into a graph that compounds over time.

Without chanio — four isolated sessions
ChatGPT, March 4
"Help me think through pricing for a healthcare platform"
Good answer. Closed the tab. Gone.
Claude, March 11
"How should I price a flat-fee vs. per-use subscription"
Bookmarked the response. Lost the bookmark.
Gemini, March 19
"Review this pricing proposal — does it hold up legally"
Didn't mention the context from March 4. Started over.
Perplexity, April 2
"Compare caregiver wage requirements in Colorado 2026"
No connection to the platform pricing question. Isolated fragment.
With chanio — one connected graph
1

Every session is a node. The March 4 pricing question links to the March 11 margin question — chanio draws the edge.

2

The April 2 research links backward to the pricing model. The graph sees the relationship; you just kept asking questions.

3

By April 15, the graph has a map of your thinking on the whole problem: what you've resolved, what's still open, what you haven't touched yet.

4

You open a new Claude session. The MCP connector loads your graph. The AI already knows you've been working on this for six weeks.

The circle and the spiral

A loop is a circle. It returns to zero.

Every AI session you run is a circle — it ends where it began. The context evaporates; tomorrow you start cold, re-explaining yourself. The topics you circle without resolving.

chanio is the one term that bends the circle into a spiral. Each session compounds onto the last instead of returning to zero — your knowledge doesn’t repeat, it climbs.

Already have months of history? /import lifts you straight onto the spiral — drop your Claude.ai or ChatGPT export and tomorrow’s brief is dense from day one.

Week 1

A few prompts. Nothing connected yet.

Month 1

Clusters form. The graph starts showing you what you keep returning to.

Month 3

Gaps visible. The graph knows what you haven't asked yet — and can tell you.

Month 6

The graph is a working model of your mind. Every new AI session starts informed.

Unlike a chat log — which is linear and flat — a knowledge graph stores relationships between ideas. The value isn’t in the individual prompts. It’s in the edges between them.

Four layers of identity, from one stream.

Your AI chats are richer than any form or profile. chanio reads that stream four ways, so the graph actually knows you.

LAYER 01

Domain

What you ask about — health, finance, music, code, relationships, the businesses you're building.

LAYER 02

Workflow

How you think through a problem — the questions you ask first, the frames you default to, the phrasings that come back.

LAYER 03

Behavioral

What pulls you back — topics you keep returning to, decisions you revisit, the stuff that won't leave you alone.

LAYER 04

Artifact

What you build from it — documents, code, plans, letters, decisions. The output of the conversation, not just the conversation.

Beyond memory

chanio doesn't just remember.
It finds your holes and your frontiers.

A graph of your prompts is also a map of what you've already thought about — and by extension, what you haven't. chanio reads that map two ways.

Gap detection

What's missing from your thinking

Topics you circle without resolving
The same question, three different ways, across six chats. chanio flags the loop so you can finally close it.
Questions you never followed up on
The thread you dropped mid-chat. The answer that raised a bigger question. chanio remembers the open ends.
Adjacent territory you haven't entered
Concepts that sit one step away from what you already think about, and that the graph says you'd probably want.
Passion expansion

Where your thinking is growing

Emerging passions
A new interest starts as three scattered prompts across a month. chanio spots the pattern before you do.
Connections across silos
Two threads you treat as separate are actually the same idea. chanio draws the edge.
The next frontier
Where the people who think like you go after they've asked what you're asking. Not recommended — inferred.

Every surface is a suggestion — never a push. You decide what to open next.

That's what the map is for. Not a record of where you've been — a way to find the next thing worth caring about.

When the question is who you are, not what's next — mapofyou reads that side of the map →

How it works

01

Capture

A browser extension sits quietly on top of Claude, ChatGPT, Gemini, and Perplexity. If you use Claude Code, chanio-core captures every session automatically at the terminal level. Either way, your prompts stay local — encrypted, never sent anywhere you haven't approved.

02

Graph

Your prompts are linked into a knowledge graph across the four layers — domain, workflow, behavioral, artifact. You see it grow in real time in your chanio dashboard.

03

Carry

An MCP connector exposes your graph to any AI that speaks the protocol. You walk into a new conversation and the AI already knows who you are — because you brought the context, not the platform.

04

Attest

When it matters — health questions, legal questions, financial questions — chanio can route the output to a signed expert. The AI drafts. A licensed human puts their name on what's true.

Why now

The infrastructure of the last era was optimized for extraction. The infrastructure of this one is optimized for attestation — and the person being attested to is you.

Compute stays rented. Storage comes home.

Compiled nightly, on your disk

Every session compiles to your local drive overnight. Not a cloud backup — a local knowledge graph that exists even when the internet doesn't. Offline-capable by design.

Owned, not rented

Your graph lives in your account and on your hardware. Export it whenever you want. Delete it whenever you want. Nobody sells it because nobody owns it but you.

Portable across every AI

MCP is the open protocol for carrying context between models. chanio speaks it natively — so the graph travels with you, not with any one platform.

Signed where it counts

For the decisions that matter, chanio can hand the output to a physician or domain expert to attest. AI drafts. Humans sign. You carry the proof.

Proof, not promise

This isn't a mockup.
It runs tonight.

The architecture on this page is the founder's daily driver, not a roadmap. chanio captures every session and compiles it overnight — a morning brief plus one living state note per venture — before 5 AM. What comes out of that job looks like this:

226
AI sessions compiled to local disk
21
venture knowledge graphs, regenerated nightly
5 AM
automatic compile while the founder sleeps
0
bytes that left the machine
chanio morning brief4:57 AM · May 20, 2026
overnight compile · 14 sessions · 3 ventures touched
Good morning. Compile finished 4:57 AM. One loop closed. One gap flagged.
venture state
chanio
ACTIVE
Closed: vault bridge wired · File System Access API
co-op.care
ACTIVE
Open: pricing page before the next partner conversation
A client proposal
CONSULTING
Status: proposal live · fixed-fee · scoped
gap detected
You've asked about “overnight compile” and “morning brief” combined 11× this month. But the product has no UI for revisiting a brief the next day.
The loop is: capture → compile → … silence. The revisit surface is unbuilt. That's the gap.
⚠ do not trust if memory says
· Flag: two pages quote different prices for the same tier → reconcile before the next deploy.
· Stale: a channel with no session in 21 days → archive it or revive it.
chanio ready · next compile 05:00 tomorrow

Compiled by chanio while the founder slept. The graph caught its own author about to quote stale pricing — and corrected him. That's the whole point: the memory you own is the memory that keeps you honest.

Try it — vault search

Ask chanio anything in your vault.

This demo runs on a redacted slice of the founder's own instance — names, dollar figures, and deal terms removed. The mechanics are exactly what runs on your disk.

This is a static slice of the chanio graph — the live product queries your own vault, compiled nightly to your local disk.

What no one else does

Other memory tools own one of these.
chanio is the only one that owns all four.

Mem, Reflect, Rewind, Notion AI, ChatGPT Memory — each gets one piece right. Verified in mid-2026: no competitor matches all four simultaneously.

Multi-provider auto-capture
Every model. One graph.

Claude, ChatGPT, Gemini — chanio captures the session structure from all three into the same schema. No competitor in 2026 captures across providers; each owns one walled garden.

Your disk, not our cloud
Markdown on your machine.

Every capture lands as a plain markdown file on your local disk — Obsidian-ready, grep-able, portable to any AI that comes next. Cancel chanio tomorrow; the corpus is yours forever.

Local-LLM synthesis
Routine work runs free.

Extraction and canonical synthesis run on your hardware via Ollama. Cloud models are reserved for the morning brief — the one step where frontier judgment compounds. ~$0.50/day total.

Per-channel canonical state
Project-aware memory.

chanio threads each project into its own channel and regenerates a 'source of truth' note per project, nightly. Open the channel and the graph already knows what's settled, what's open, what was last touched.

One graph. Every surface. Yours on disk.

Why it matters

Everyone will have the tools.
What's yours is the difference.

Intelligence is becoming a utility — metered, the same for everyone. What isn't a utility is you: your context, your judgment, the years of thinking only you have done. chanio keeps that on your side of the line — so the AI amplifies you instead of replacing you.

The model is rented. The mind is yours.

Stop renting your identity.
Start owning your context.

Join the waitlist for the private beta. You'll get early access, a say in the roadmap, and locked-in founding pricing when we open the doors.

We store your contact only to invite you. No public listing.