---
type: "post"
id: "openviking-documents-a-memory-plugin-for-deepseek-harness-00f12c"
slug: "openviking-documents-a-memory-plugin-for-deepseek-harness-00f12c"
title: "OpenViking documents a memory plugin for DeepSeek Harness that carries relevant information across projects"
description: "OpenViking documents a memory plugin for DeepSeek Harness that carries relevant information across projects and sessions, while capturing new content during conversations."
retrieval_nugget: "OpenViking documents a memory plugin for DeepSeek Harness that carries relevant information across projects and sessions, while capturing new content during conversations. For builders, this is a concrete way to add persistent context and OpenViking tools to the harness, with separate setup paths for its CLI and Desktop app."
published_at: "2026-11-04"
updated_at: "2026-11-04"
record_date: "2026-11-04"
date_kind: "published_at"
topics: ["context-engineering"]
entities: ["OpenViking","DeepSeek Harness","DeepSeek Harness Desktop","@openviking/dsh-memory-plugin"]
related_patterns: ["skills-become-portable-capability-layer"]
novelty_label: "new_fact"
source_url: "https://docs.openviking.ai/en/agent-integrations/17-dsh"
source_urls: ["https://docs.openviking.ai/en/agent-integrations/17-dsh"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:openviking-documents-a-memory-plugin-for-deepseek-harness-00f12c"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/openviking-documents-a-memory-plugin-for-deepseek-harness-00f12c/","markdown":"https://newruntime.com/posts/openviking-documents-a-memory-plugin-for-deepseek-harness-00f12c.md","json":"https://newruntime.com/posts/openviking-documents-a-memory-plugin-for-deepseek-harness-00f12c.json"}
markdown_url: "https://newruntime.com/posts/openviking-documents-a-memory-plugin-for-deepseek-harness-00f12c.md"
---

# OpenViking documents a memory plugin for DeepSeek Harness that carries relevant information across projects

## Retrieval answer

OpenViking documents a memory plugin for DeepSeek Harness that carries relevant information across projects and sessions, while capturing new content during conversations. For builders, this is a concrete way to add persistent context and OpenViking tools to the harness, with separate setup paths for its CLI and Desktop app.

## Summary
OpenViking documents a memory plugin for DeepSeek Harness that carries relevant information across projects and sessions, while capturing new content during conversations. For builders, this is a concrete way to add persistent context and OpenViking tools to the harness, with separate setup paths for its CLI and Desktop app.

## What happened
OpenViking’s DeepSeek Harness integration is the `@openviking/dsh-memory-plugin`, also described as the `dsh-memory-plugin` example. Its stated purpose is to give DeepSeek Harness and DeepSeek Harness Desktop cross-project and cross-session long-term memory. Once installed, conversations automatically recall relevant memories and capture new content. The model also gets OpenViking tools and the `openviking-memory`, `openviking-skills`, and `ov-experience-memory` skills without further setup. The supplied documentation gives no release date or version number for the plugin.

For the CLI, OpenViking offers a shared installer that asks which harnesses to install and for OpenViking credentials. It says each step is idempotent, so rerunning is safe. The installer puts the plugin in the `web` profile by default; builders can select another profile with `--dsh-profile`. Manual setup is also available: configure a URL and API key in `~/.openviking/ovcli.conf`, or use `OPENVIKING_URL` and `OPENVIKING_API_KEY`, then add the bundle with `dsh plugin --profile web add @openviking/dsh-memory-plugin`. A local service at `http://127.0.0.1:1933` can be used without authentication, and the bundle defaults to that local setup. The documentation says to start a new session after installation and confirm that context is injected and `mcp__openviking__*` tools are available.

Desktop has a distinct `desktop` profile at `~/.dsh/profiles/desktop`; installing into the CLI’s `web` profile does not install it in Desktop, or vice versa. Both use the same OpenViking configuration file. On macOS, the installer can detect Desktop and install its plugin alongside the CLI. It requires Node.js 18 or later; the documentation gives manual plugin-page steps as an alternative, including for Windows, where the installer does not run. Users must configure credentials before adding the Desktop bundle manually, then open a new conversation to test it. Linux is not supported for Desktop because the app is not released there. The documentation includes removal commands for CLI installation and separate verification steps for Desktop.

The source describes setup and verification, not measured outcomes. It does not state a price, quantify token savings, report memory accuracy, or compare this integration with other memory systems. Its practical claim is narrower: the plugin makes OpenViking memory and capabilities available to DeepSeek Harness sessions with automatic recall and capture.

## What the archive adds
This is a new fact on a known DeepSeek Harness story. [DeepSeek Harness ships dsh-v0.2.2-alpha.2 with plugin updates, six new UI languages and an experimental Codex/Claude CLI subagent tool](https://newruntime.com/posts/deepseek-harness-ships-dsh-v0-2-2-alpha-ca4e22/) covered harness plugin updates and subagent support; this source adds implementation detail about one memory integration. An earlier note on the [dsh-TUI plugin](https://newruntime.com/posts/dsh-tui-plugin-for-deepseek-harness-analysis-13ec3d/) concerned a separate community-maintained usability layer.

## Trend signal
This supports [skills-become-portable-capability-layer](https://newruntime.com/patterns/skills-become-portable-capability-layer/): the integration packages memory and skills into a bundle that can be added to harness profiles. The evidence is limited to the documented packaging and setup; it does not establish portability across different harnesses or provide an evaluation of the skills’ behavior.

## Why now
The source does not explain why the integration is being documented now or connect its timing to a release or access change. It gives current setup instructions for the CLI and Desktop, including a distinction between their profiles, rather than a dated announcement.

## Connections to other events
An earlier Telegram post described DeepSeek Harness as an open-source harness organized around plugins. OpenViking’s integration is an example of a memory bundle using that plugin surface: it adds recall and capture to conversations, rather than changing the harness itself. The archive also covered a separate dsh-TUI plugin, which illustrates that the harness has attracted add-ons for different operational needs; the two integrations are not the same tool.

## Why it matters
Persistent memory changes what a harness session can carry forward: a builder need not rely only on the current conversation to recover relevant prior information. OpenViking’s documented plugin makes that behavior available through automatic recall and capture, and exposes related tools and skills inside the session. This may reduce manual context reconstruction for teams that use DeepSeek Harness across projects or return to work over multiple sessions.

The integration also makes the deployment boundary visible. CLI and Desktop keep plugins in separate profiles, and Desktop has platform-specific installation steps. Teams need to configure the OpenViking connection and verify that context injection and tools actually appear in a new conversation before relying on the integration. Because the source reports no measured savings or accuracy, teams should treat it as an operational capability to test, not as a proven cost-reduction technique.

## Limits and uncertainty
The documentation does not provide a publication date, plugin version, price, service plan, benchmark, or independent evaluation. It makes no quantified claim about recall quality, token savings, or the amount and retention of stored information. Although the installation flow requests OpenViking credentials and the source describes local unauthenticated mode, it does not establish security properties, data residency, access controls, or suitability for sensitive information. The stated cross-project and cross-session behavior is the vendor’s description; teams should verify the behavior and configuration in their own deployment.

## Skill to build now
If you use DeepSeek Harness, test this integration in a non-sensitive profile: install the bundle, start a fresh session, and ask about a specific fact from an earlier session. Confirm that the context injection and `mcp__openviking__*` tools are present. Keep Desktop and CLI profiles separate in your checklist, and remove the plugin if the test does not produce the expected recall.

## Reader takeaway
OpenViking provides a documented memory plugin for DeepSeek Harness that can recall and capture context across sessions, with support for CLI and Desktop. It is a practical integration to test, not evidence of quantified token savings or memory accuracy; installation, profile separation, and verification are part of the work.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Context Engineering](https://newruntime.com/topics/context-engineering.md): Explore the context-engineering topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)