Kazn · How it works
One memory. Every agent. You approve.
Kazn sits between you and the AI agents you already use. They all read the same project memory, they propose what to add or change, and you decide what becomes true.
Your AI agents
Kazn
Project Brain- What is this?
- Why does it exist?
- What’s decided?
- Where does it stand?
- What happens next?
One address for every agent. Access per project. Every call recorded.
You
Step by step
From connection to shared knowledge.
- 01
Connect your agents
Give each AI agent one Kazn address and its own key. Choose which projects it can read, and whether it can propose changes.
- 02
Every agent starts informed
Each project has a living Project Brain: what it is, why it exists, what’s decided, where it stands and what happens next. A newly connected agent reads it first.
- 03
Agents propose, you approve
When an agent wants to record a decision or change a page, the proposal lands in your Inbox beside the current version, with its reason. Approve or reject it, on the web or on your iPhone.
- 04
Everyone knows at once
An approved decision is searchable almost at once, by meaning as well as by words. The next agent you open already knows it, and the replaced decision stays in the history.
- 05
Everything is recorded
Every proposal, approval, edit and agent read is in Activity, so you can always see who changed what, when and why.
In practice
Three everyday moments.
Switch agents without starting over
You plan a feature with a chat assistant in the morning and build it with a coding agent in the afternoon. The coding agent reads the decisions you made in the morning; you don’t explain them again.
Change your mind safely
You decide to move from one database to another. An agent proposes the new decision, you approve it, and the old one is marked replaced and linked. Every agent now gets the new answer, and the reasoning behind both stays readable.
Keep an agent in its lane
A research agent can read one project and nothing else. When it tries to read another, Kazn refuses and records the attempt, and you see it in Activity.
Works with
The agents you already use.
Any AI agent that supports the Model Context Protocol (MCP), the open standard for connecting agents to tools and data. Kazn’s own demo runs with Claude Code and Codex.
Add MCP servers once in Kazn and choose which agents may use their tools, instead of setting up every tool in every agent.
- WebThe full workspace in your browser.Early access
- iPhoneA native app, coming first on TestFlight.In progress
- Kazn assistantAsk in your own words, on the web and iPhone.In progress
- AndroidA native app on the same Second Brain.Planned
- DesktopA desktop app on the same Second Brain.Planned
Get invited as places open.
Kazn is in early access. Join the list and we’ll invite you as places open.
Join early access