This doesn't land at first. "Why split it up. Isn't one memory enough?" But once a little piles up, memory you didn't split becomes unusable. Two things are all you need: who owns the memory (the team) and how you get into it (the project).
Why split: what happens if you don't
Pour every memory into one bucket and:
- Work and your personal fiction drafts come back in the same search.
- You want to show a colleague a work memory, and your private notes open too.
- You ask about one project's decision and another project's tags along.
Memory is useful inside a context. Mix it and that context collapses. So AiAkiv has teams own memory, and splits access through projects.
Team: the owner of the memory
- Memory belongs to a team. You need a team to save or read at all.
- Signing up creates a personal team automatically: usable right away.
- You can make as many more teams as you like, and invite other people to turn one into shared memory.
- Sharing, isolation, and billing all draw their boundary at the team. One team's memory does not leak outside that team.
Project: the way into that memory
- A project is a way into the team's memory. Not a separate warehouse, but a door into memory the team already has.
- One team memory can have several projects. Each way in is configured
differently to control saving precisely:
- Sub-folder: the compartment memories saved through this way in land in.
- Persona: the role and voice the AI takes on this way in → Saving.
- The way in that saving currently points at is called the Main project.
As a picture
Account (email sign-in)
└─ Team = owner of the memory (personal team on signup · shared by invitation)
└─ Team memory ←── several Projects (ways in)
Project = tunes saving via (sub-folder · persona)
The one saving goes through now = Main
If you're using it lightly on your own, a personal team plus the default project is a fine start. When you later want to split things up, make more projects, or make a new team to share with someone.
The Main project: where saving goes right now
A memory landing in the wrong way in is the number-one accident for newcomers. So the current target is always checkable.
- The check target tool tells you the current account, team, project (Main), sub-folder, and the switch link.
- Search and save responses also show this information every time, the save location is always visible.
- One exception, folder binding: an AI that uses folder-based MCP (Claude CLI, for instance) can pin a project to a folder (folder-project binding). Then you don't have to think about Main at all. Setup in Folder binding.
- Main can be turned off for a while. Press "Turn Main off" in the console's project management and, until you turn it back on, connections that follow Main (account connectors, API keys not pinned to a project) refuse every request, while folder connections and project-pinned API keys keep working. It suits someone who uses only folder connections on a PC and only Main on a phone → When the account connector and a folder connection are both attached.
Only people switch it: the guard
By default, only a person changes the save target. This is the guard that stops prompt injection from quietly changing where your data gets written.
- The check tool is read-only: it shows the location, it can't change it. The save and search tools can't change it either.
- To change it, a person clicks the console's switch link (
switch_url). - The exception is one you open yourself. Turn on the save target switch permission in the console's Reference → AI permissions and you can ask the AI "switch my save target to project X". It starts off, and while it is on the console's top bar shows a notice. A connection made with an API key also needs that key to carry the manage scope (Authentication). The destination can only be one of your projects or a listed public target, and connections pinned to a folder or a project key do not move → Asking your AI to do console work. Asking the AI works only on a connection that has app features; on one that does not, you switch it in the console yourself → The app action tool.
- While Main is turned off, the AI can't change or clear it even with this permission.
- Turning Main off and on is also a person's job by default. With "Turn Main off" and "Turn Main back on" on in AI permissions, you can ask the AI. They are turned on separately, so you can leave turning it off to the AI and turn it back on yourself. The AI can turn Main on only from a connection pinned to a folder or a project API key (while Main is off, account connectors refuse every request) → Asking your AI to do console work.
Situations that come up
- "Saving goes somewhere unexpected." → Check with the check-target tool first. If it isn't what you expected, switch in the console. (With folder binding you never hit this → Folder binding.)
- "A memory I expected isn't there." → Before assuming data loss, check whether the team and project match between when you saved and when you're searching. Most of the time it isn't loss, it's a target mismatch → Common problems.
Next
- What you may store (and security) → What to store
- Using it together inside a team → Team sharing