Plugin DSH
简体中文

Memory plugins for DeepSeek Harness

Carry context across sessions so the agent stops forgetting.

Memory plugins keep information around after a session ends, so the next one does not start from nothing. Some store short notes about you and your project, some summarise past conversations, and some save everything and retrieve the relevant parts on demand. The shared goal is to stop you repeating the same background every time.

2 plugins

How to choose

Find out where the memory is stored and whether you can read and edit it; plain files on your own disk are the easiest to inspect and correct. Check how much gets injected into each request, since a plugin that recalls too much costs tokens and can crowd out the task at hand. Running two memory plugins together often produces duplicate or conflicting recall, so start with one.

Terms used on this page

Context window
The amount of text a model can take in for one request; memory has to fit inside it alongside everything else.
Recall
The step where stored memories are selected and added to the current conversation.
Compaction
Shortening older conversation into a summary so the important parts survive within the context limit.

Questions and answers

Where do memory plugins keep my data?

It depends on the plugin: local files, a local database, or a remote service. The readme should state this plainly, and if it does not, treat that as a reason to look at the source before installing.

What if the agent remembers something wrong?

A wrong memory keeps coming back until it is removed. Prefer plugins that let you view, edit and delete individual entries, and review what has been stored from time to time.

Arrow keys to move, Enter to open, Esc to close