Does chanio store my AI conversations in the cloud?

No. chanio runs no content servers -- your AI sessions are captured and compiled to plain markdown on your own disk. The only cloud calls are optional, for the single step where frontier judgment is needed, and you bring your own key and cap your spend.

Do I need to install anything or get an API key to try chanio?

No. You can see a sample capture in seconds with no signup, and chanio works with the AI you already pay for, or free on iPhone with Apple Intelligence. An API key is only needed if you want chanio to run the model itself.

What happens to my existing ChatGPT or Claude conversation history?

The export-rescue tool at chanio.com/export-rescue parses your export ZIP entirely in the browser -- no upload, no account, no API key -- into a knowledge-graph preview, and one click downloads the whole thing as an Obsidian vault of wikilinked markdown.

SET UP THE REAL THING

A real model. A real memory. Your Mac.

The in-browser demo is the spark — this is the fire: a genuine local model (via Ollama) answering from a memory folder on your own disk. Plain markdown, so it's Obsidian-compatible out of the box. Ten minutes, free forever, nothing leaves your machine.

The easy way — your AI builds it for you

Open the AI you already use (Claude, ChatGPT, Gemini — any of them) and paste this one prompt. It installs everything and explains each step. That's the whole setup — which is very much the point.

paste into your AI

Set up "chanio-lite," my own private AI brain, on my Mac. Do it step by step and explain as you go:

1. Check if Ollama is installed (run: ollama --version). If not, install it with: brew install ollama — or tell me to download it from ollama.com/download.
2. Start Ollama and pull a small model that runs fine on a laptop: ollama pull llama3.2:3b
   (If my Mac has 16GB+ RAM and I want a stronger brain, also: ollama pull qwen3:14b)
3. Make a folder ~/chanio/notes/ — I'll drop .md or .txt notes in there (my plans, ideas, anything I want it to know). It's plain markdown, so I can open the folder in Obsidian too.
4. Create ~/chanio/ask.mjs: it takes my question from the command line, keyword-searches my notes for the relevant lines, and sends those lines + my question to the local model at http://localhost:11434/api/chat (model llama3.2:3b, stream:false, think:false). It prints a 2-4 sentence answer. Then it makes a SECOND quick local call asking the model to judge — as JSON {"escalate": true|false} — whether the question is meaningful enough for a frontier AI (a hard decision or deep reasoning; most are NOT). If true, it appends the question as a line to ~/chanio/todays-questions.md.
5. Create ~/chanio/tonight.mjs: it prints everything in todays-questions.md as one clean prompt for me to paste into the AI I already use (with a note asking each AI to answer thoughtfully), then archives it to ~/chanio/archive/<date>.md and clears the file.
6. Create ~/chanio/growth.mjs: it reads ~/chanio/feedback.md (where I paste the AIs' answers back), sends it to the local model asking for exactly three sections — CONSENSUS, DISAGREEMENT (and which view is more convincing), and 3 TAKEAWAYS — then saves the result to ~/chanio/growth/<date>.md and clears feedback.md.
7. Use only Node's built-in modules (no npm install). When done, show me the three commands:
   node ~/chanio/ask.mjs "my question"
   node ~/chanio/tonight.mjs
   node ~/chanio/growth.mjs

Keep everything local and simple. Nothing should call any cloud API.

What you end up with

A real brain

llama3.2:3b runs on any modern Mac; got 16GB+ RAM? Pull qwen3:14b for a noticeably deeper one — chanio's own founder runs the Qwen family.

A real memory

~/chanio/notes — plain .md files on your disk. Drop anything in; open the folder in Obsidian and it's a vault. The model answers FROM it.

The whole loop

ask (instant, free, private) → tonight (the day's big questions, packed for the frontier AI you already pay for) → growth (what all the answers taught you, synthesized locally).

Prefer it done by hand?

1. Install Ollama — brew install ollama (or ollama.com/download)
2. ollama pull llama3.2:3b — and optionally qwen3:14b for the bigger brain
3. Make ~/chanio/notes/ and drop in a few .md notes
4. Paste the prompt above into any AI to generate the three small scripts — or write them yourself; they're each under a page of plain Node.

Windows & Android: the paste-at-night ritual works everywhere today; the full local setup lands 2027. iPhone & iPad: soon — today the in-browser model at /try runs on modern iPhones.

Join the beta — get the full loop run for you

or start with the in-browser model right now →