It started as an accident.
I’ve been using Krisp and other services to automatically take meeting notes for a couple of years now, and it’s incredibly valuable. But I didn’t want my data locked in that ecosystem. I wanted to use meeting notes alongside handwritten notes, emails, tasks, and everything else - Slack, Notion, iMessage, hand taken notes - all connected so I could actually figure stuff out quickly.
There’s nothing worse than knowing exactly what you want - just find that link to update the particular Slack canvas someone asked about - and spending fifteen minutes digging through six different solutions trying to find it - was it an email? A Slack message? A text?
That’s broken.
So I started building KAIPARR. Just for me. One of those personal efficiency projects you hack together because you’re tired of the friction.
But somewhere along the way, I realized something. I’ve been working in AI since it became practically and economically feasible in the 2010s, and the entire conversation has become… narrow. We talk about AI as a tool to automate white-collar work, to make developers more productive, to eliminate jobs altogether - which I guess is some people’s dream. But we’ve forgotten the actual point.
This technology, like every technology, is only valuable to the extent that it actually makes humans’ lives better.
AI Worth Building
I don’t see how making AI something that eliminates jobs or reduces the need for people to work is valuable. A bunch of humans without work isn’t a win.
But I do see genuine value in something different: connecting the data that matters to you, answering your questions, and having the system perform tasks on your behalf. Draft responses to important emails. Create tasks from emails received overnight. Take that incredibly long thread that’s been sitting in your inbox - the one about scheduling a play date - and actually resolve it instead of it sitting there draining your mental energy.
That’s meaningful. Reducing cognitive load for real people doing real work. That’s what AI should do.
The Architecture
Underneath KAIPARR is an AI harness that connects to any MCP-enabled service - Notion, GitHub, Slack, and hundreds of others. The same harness connects to your local machine so it can send emails, navigate your already-logged-in browser, check account balances, view and edit your documents, or handle other tasks you permit it to do.
Think of it as a shell. Its job is to connect the data and information that matters to you and enable AI to work on your behalf.
Three functional pillars:
- Act on permitted tasks without you having to coordinate across six different applications
- Provide the right context for decision-making by connecting everything that’s relevant
- Reduce the daily barrage of information so you can actually think
Built for Real People, Not Enterprises
This isn’t designed as an enterprise automation tool to help companies reduce headcount. It’s designed for regular people with busy work and family lives. People trying to stay organized, delegate what can be delegated, and reclaim time.
Simple, high-friction tasks. Tracking a package. Finding a link. The cognitive load of context switching between six tools to accomplish something that should take thirty seconds.
KAIPARR is meant to eliminate that.
Privacy and Control
Here’s what matters: the only data stored inside KAIPARR is what you put there directly - manual notes, tasks if you choose to manage them there, and your chat history with the AI.
Chat history is retained specifically to give the system context. It learns your preferences, your writing voice, how you communicate - so when it drafts an email, it sounds like you. It remembers who you talk to frequently. That’s the point of memory.
But all of that data is private, secure, and owned by you. Not used for training. Not monetized.
Here’s the part that actually matters for privacy: you supply your own API key to Anthropic (or whoever your AI provider is). The user owns the AI account relationship - not KAIPARR. That means you control what data goes to the AI provider. You can instruct Anthropic directly to opt out of training data usage. You own the full privacy chain.
We’re planning to add Gemini and OpenAI as options too, so you can choose which models and services you work with. Your choice. Your control.
KAIPARR doesn’t get between you and your AI provider. We’re a layer that helps you use it better.
What This Actually Is
KAIPARR is a personal assistant for people who are drowning in tool fragmentation and information overload. Not a replacement for humans. Not a job elimination machine. Just a way to reclaim focus and time by connecting the stuff that matters and letting AI handle the coordination.
It’s the opposite of what most AI products are trying to do.
And honestly, that’s the point.
— Brian Roy
Originally written as spoken notes for KAIPARR
April 15, 2026