eigent/docs/agents/memory.md

---
title: Agent Memory
description: Understand the context Eigent uses today and the planned direction for persistent agent memory.
icon: brain
---

The Agent Memory page is present in the Agents dashboard but is currently marked **Coming soon**.

<Warning>
Do not rely on this page for persistent cross-project memory. The feature is not generally available in the current product.
</Warning>

## What Eigent remembers today

Current context can come from:

- The active Project conversation
- Earlier Runs in the same Project
- Files in the active Space
- Uploaded task attachments
- Project instructions, rules, and tone
- Enabled Agent Skills
- Tool and connector configuration

These sources are different from a dedicated long-term memory system.

## Expected memory controls

The final feature design is not yet available. Persistent memory is expected to
include controls for:

- What information an agent can store
- Whether memory is scoped to a user, Space, Project, or agent
- How a memory is created
- How users review and edit memories
- How users delete or disable memories
- Retention and synchronization behavior
- Sensitive-data controls
- How memory affects prompts and model usage

> **Screenshot placeholder:** Add a screenshot only after the Memory page contains active controls. Do not use the current Coming soon screen as the primary product image.

> **Video placeholder:** Add a video only after users can create, review, and delete a memory. The video should include all three actions and explain scope.

## Use current alternatives

Until Agent Memory is available:

- Put stable project behavior in project instructions.
- Put reusable expertise in Agent Skills.
- Keep related follow-ups in one Project.
- Store durable reference material in the Space workspace.

## Related guides

- [Workspace instructions](/projects/workspace)
- [Agent Skills](/core/agent-skills)
- [Project runs](/core/project-runs)