> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orbitagents.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Agents that learn

> Your agents get better every week, recording what worked and how you like things done.

Agents in Orbit do not start from scratch every run. They record what worked and how you want things done, and they review their own recent work on a regular beat, so the operation gets sharper over time instead of resetting.

## How the learning happens

* **Agents record learnings** as they work: what succeeded, and the preferences you corrected them on.
* **Every 10 runs or 7 days**, an agent reviews its recent work and pulls out what to keep doing.
* **Learnings have a scope:** kept to one agent, shared with a project, a department, or the whole workspace. A lesson one agent learns can lift the whole team.

## Agents can plan their own schedule

Ask in chat and an agent will schedule itself ("I will run every Monday at 9") or set up another agent to cover a gap. New agents are only ever created with your yes.

## Why this matters

This is the difference between a tool and a teammate. A tool does the same thing every time. An Orbit agent remembers the win, the correction, and the preference, and applies it next time, so your company's know-how compounds in the [Brain](https://docs.orbitagents.com/concepts/knowledge.md) rather than living in your head.

> [card] [How shared memory works](https://docs.orbitagents.com/concepts/knowledge.md)
>
> The one Brain every agent reads and writes to.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.