AI Agent Hub
Back to plugins
🧠

dsh-shared-memory

Memory Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install futongxu9-maker/dsh-shared-memory

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install futongxu9-maker/dsh-shared-memory in the DeepSeek Harness terminal, or fetch the source from https://github.com/futongxu9-maker/dsh-shared-memory and follow the manual install steps in the README.

About this plugin

Every new conversation starts from a blank slate. The preferences you corrected last session, the project paths you spelled out, the working principles you settled on — all of it is lost. dsh-shared-memory closes this gap with layered persistent memory: MEMORY.md and USER.md are injected into every new session's system prompt, while files under the topics directory stay out by default, represented only by a one-line index entry so the agent can pull the full content on demand.

Beyond automatic injection, the plugin ships memory maintenance tools and a visual side panel. The tools support add, replace, remove, and list operations targeting memory, user, or any topic file, with a built-in self-check rule that only permits persistent principles, work data, and lasting facts — rejecting one-off or ephemeral content. The panel organizes the three memory lines (notes, profile, topic library) across three tabs, making browsing, editing, and deletion straightforward, with a live capacity indicator.

It is well suited for users who run long-running, multi-project sessions in DeepSeek Harness and want the agent to retain style, environment, and context across conversations — so memory actually carries forward instead of resetting with every new chat.

Use Cases

  • Maintain context and preferences across long multi-project sessions
  • New conversations inherit user profile and work data without re-explanation
  • Load topic files on demand to keep the system prompt bounded

Best For

  • DeepSeek Harness users running long multi-project sessions
  • Developers who want the agent to retain style and context across chats
  • Team members who prefer durable facts over repeating instructions each session