Use a brain as agent memory
Have your AI tool read, save and search what it learns in a shared brain.
A brain works as long-term memory for an AI tool: it reads what's there before it starts, saves what it learns, and finds it again later. Because the brain is shared, every teammate's agent works from the same pages.
Before you start
- An account and a brain. See Sign up and create a brain.
- Your AI tool connected to the MCP server. See Connect over MCP.
Ask your AI tool to use the brain
Name the brain in your request. These examples use a brain called example-project:
| You ask | The tool calls |
|---|---|
| "List my OpenKnowledge projects." | projects, which shows example-project |
| "Read notes/about.md in example-project and summarize it." | exec with cat notes/about.md |
| "Save what we decided today to decisions/2026-11-10.md in example-project." | exec, which writes a new page |
| "Search example-project for anything about onboarding." | search |
| "What links to decisions/2026-11-10.md?" | links |
Make it a habit
Most AI tools let you give standing instructions, such as a project's CLAUDE.md or AGENTS.md, or custom instructions in Claude and ChatGPT. Add a line like this so the tool uses the brain without being asked:
Our team's notes live in the OpenKnowledge brain example-project. Check it before answering questions about our work, and save decisions and findings there.See what the agent changed
Open the brain in the app. Each page's version history shows the agent's changes, credited to the agent acting for you, and you can restore an earlier version.