> For the complete documentation index, see [llms.txt](https://docs.lleverage.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.lleverage.ai/operator-agent/memory.md). # Memory Memory is what lets the agent learn. As it works through sessions, reads files, and receives feedback, the agent stores what it learns as memory - and reads it back in later sessions. That means it carries context between conversations: tell it something once, and it remembers next time. Memory is available on selected plans. ### How Memory Works During a session, the agent reads, writes, and searches its memory automatically. You don't need to manage this - it happens as part of the agent's normal work. Memory matures through three layers: * **Observations** - Raw and of-the-moment. Past sessions, reflections, and ingested files. * **Memories** - Distilled and reusable. Facts, preferences, and rules of thumb. * **Knowledge** - Settled reference. Curated articles, skills, and rules. New information enters at the bottom as observations, gets distilled into reusable memories, and can eventually settle into curated knowledge the agent treats as authoritative reference. ### The Memory Explorer Open **Intelligence > Memory** in your project to see everything the agent knows. The page shows memory as a layered map - three stacked planes, one per layer, with each document as a node. Related memories cluster together, and lines connect memories to the sessions and files they came from, so you can trace where any piece of knowledge originated. The map is live: when the agent stores or updates something during a session, the map refreshes on its own. To find something specific: 1. **Use the search field** in the side panel to search across all of memory. 2. **Select a result** to open the document and see exactly what the agent has stored. 3. **Click any node on the map** to preview that document in place. You can zoom, rotate, and reset the view using the on-screen controls, and the whole map can be navigated with the keyboard. Summary counts show how much sits in each layer. ### Reading and Editing Memory Anyone with access to the page can read memory documents. Each document shows its content alongside its layer, its type (for example a fact, preference, reference, or article), and when it was last updated. Users with the Builder role can also change what the agent knows: 1. **Open a document** from the map or a search result. 2. **Select the edit control** to switch the document into editing mode. Documents are plain markdown. 3. **Make your changes**, then choose **Save** - or **Discard changes** to leave the document as it was. Builders can also delete documents the agent should no longer rely on. > 💡 Editing memory directly changes how the agent behaves. If it has remembered something wrongly - an outdated fact, a preference that's changed - correcting the document here fixes it for every future session. > ⚠️ Deleting a memory removes it from what the agent knows. If you only want to correct a detail, edit the document instead. ### Seeding Memory in a New Project When a project has no memory yet, the Memory page helps you give the agent a head start. You can: 1. **Run a background check** - the agent researches your company and stores what it finds. 2. **Connect tools** - link the apps you work in so the agent can learn from them. 3. **Upload files** - share documents, decks, or notes for the agent to learn from. As each step completes, you'll see the first memories appear on a live map. You only need one of these to get going - the agent keeps learning from every session after that. ### Memory in Chat While you chat with the agent, a **Memory** card in the sidebar shows a miniature version of the memory map. When the agent stores something new mid-session, the card animates - zooming in on the new memory so you can see the moment it learns. Select a layer to list its documents, open any of them to read what was stored, or follow the link to the full Memory page. > 💡 The more the agent knows, the more useful it is. Mention your preferences, correct it when it gets something wrong, and give it feedback - all of that becomes memory it applies in future sessions. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.lleverage.ai/operator-agent/memory.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.