> For the complete documentation index, see [llms.txt](https://docs.zerve.ai/zerve-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.zerve.ai/zerve-docs/superagents/memory.md).

# Memory

How a superagent remembers decisions, findings and your preferences over time.

A superagent remembers what it has learned and decided, across conversations and over weeks of work. Its memory is what lets it build on earlier work instead of starting again, and remember how you like things done.

## What it remembers

A superagent's memory is a set of short notes, each of one kind:

| Kind           | What it records                                                                    | Example                                                              |
| -------------- | ---------------------------------------------------------------------------------- | -------------------------------------------------------------------- |
| **Decision**   | A decision it made                                                                 | "Use daily rather than intraday data for the signal tests"           |
| **Finding**    | Something it established                                                           | "The value factor stopped working in European small caps after 2021" |
| **Preference** | Something you prefer                                                               | "Report results in basis points, not percentages"                    |
| **Constraint** | A limit it must respect                                                            | "Never use data after the 15th of the month in month-end tests"      |
| **Context**    | Useful background                                                                  | "The risk team reviews exposures every Monday morning"               |
| **Pattern**    | A playbook for a recurring kind of work: what worked, the pitfalls and what failed | "How we build and validate a new factor"                             |

Patterns are especially useful: before starting similar work again, the superagent checks its patterns, so it reuses what worked instead of working it out again.

## How memory is kept accurate

Each note records where it came from, such as the conversation, team member or task, and can keep the exact words it was based on, such as what you said. When a note becomes out of date, the superagent replaces it with a new one and links the two, rather than editing the original, so there's a record of what changed and why. Notes that no longer matter are retired.

The superagent brings its most relevant active notes into every turn of work.

## Telling it what to remember

You don't need to do anything special. When you state a preference or a constraint in a conversation, such as "always exclude penny stocks," the superagent records it. You can also ask it directly:

> Remember that the investment committee wants every recommendation to include a downside scenario.

> What do you remember about how I like the weekly report formatted?

## Memory and Learnings

Superagents have two kinds of memory, at different levels:

|                | Superagent memory                                                                 | [Learnings](/zerve-docs/agents-in-the-workspace/learnings.md)                                         |
| -------------- | --------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- |
| **Belongs to** | The superagent and its coordinators                                               | Each worker's notebook                                                                                |
| **Holds**      | Decisions, findings, preferences, constraints and playbooks for the whole mission | The detailed research behind one piece of work: objectives, assumptions, approach and research topics |
| **Shared**     | Across the superagent's leadership, in every conversation                         | Within that notebook                                                                                  |

When the superagent gives a worker new work, it can pass on what it knows as starting **learnings**, so the worker doesn't start from scratch. These are only added when the worker has no learnings yet, so they never overwrite what the worker has learned since.


---

# 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 by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.zerve.ai/zerve-docs/superagents/memory.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

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.