Does turning Knowledge off delete what an agent wrote?

No. The per-chat toggle just skips the Knowledge Base for that one conversation, and the per-agent switch stops the agent from using it. The notes stay in the workspace either way.

Do agents share memory with each other?

Yes, by design. Every agent in a workspace reads and writes the same Knowledge Base, so what one agent learns, the others can use.

Who can see what an agent remembers?

Everyone who is a member of the workspace. The Knowledge Base is a shared surface, which is exactly why agents are told to keep it impersonal and workspace-relevant.

Agents remember with Knowledge

Agents do not keep a private memory: they remember through the workspace Knowledge Base. Anything an agent learns, it writes down as a note, and anything you write there, every agent can read 🧠

The always-injected Index

At the start of every run, the agent reads the Index note of the workspace Knowledge Base. It works as a map of what is known and where it lives, so the agent can find the right note without being told. Keep the Index current and every future conversation starts oriented.

The agent is instructed to treat the Knowledge Base as its persistent memory: always write durable information it learns, such as decisions, conventions, how-tos, project facts, user preferences, and operating notes, and never keep important state only in the conversation.

Agents write what they learn

You can ask an agent to save something directly: "write this to Knowledge" is enough. The agent upserts notes by title, so writing again to the same title updates the note instead of duplicating it.

The Knowledge Base is shared with everyone in the workspace. Keep it impersonal and workspace-relevant. Never store personal or colleague-related content there.

Turn Knowledge off for a chat

Sometimes you want a clean slate: a one-off conversation where the agent neither reads the Knowledge Base nor writes anything new. You can switch Knowledge off for a single chat without touching any notes.

Open the tools dropdown in the composer, the sliders icon next to where you type.
Toggle Knowledge off. A check mark means it is on; an × means it is off for this chat.
Send your message. The agent answers without touching the Knowledge Base.

The toggle is per chat, so it only affects the conversation you are in. Your other chats keep Knowledge on.

Turn Knowledge on or off for an agent

Each agent has a Knowledge switch in its settings. Turn it off and the agent stops using the workspace Knowledge Base entirely. Turn it back on and the notes are all still there, so toggling it off does not erase anything.

Next to the switch is an Open Knowledge button. Open it to read what the agent has written, edit a fact that is wrong, or delete something stale.

Use Knowledge from outside idapt

The Knowledge Base is available beyond the app. Through the idapt API, SDK, CLI, and MCP, an external agent can write to and read from the same workspace Knowledge Base. That lets a coding agent in your terminal share what it learns with every other agent in the workspace. See Connect any coding agent.

FAQs

What is the Knowledge Base?

One Knowledge Base per workspace: markdown notes with folders, wikilinks, tags, and a graph, shared by your team and your agents.

Agents

Create AI personas with custom instructions, shared Knowledge, and their own default model.

Writing together

Edit the same note with teammates in real time, and let your agents read and write the Knowledge Base too.

Was this helpful?