dsh-env-inspector: Environment Self-Check and Port Management
`dsh-env-inspector` is an environment self-inspection plugin for DeepSeek Harness, providing a visual dashboard to manage system status. Core features include: real-time port detection and safe release (with a core port protection mechanism), displaying system and CLI toolchain status, globally detecting large-model API keys, listing plugin versions, and distinguishing between LAN and public IP addresses. The plugin has zero runtime dependencies, performs local self-inspection without internet access, and strictly protects privacy while preventing plaintext API key leakage.
Read MoreDeepSeek Harness 3D Pig Trough Fish (Whale) Desktop Pet Plugin
DeepSeek Harness 3D Pig Trough Fish (Whale) Desktop Pet Plugin, maintained by developer rongzi5, is designed to provide a visual interaction carrier for DSH agents. It integrates LLM chat, memory management, session progress queries, and sub-agent task scheduling. Key features include Three.js-based 3D rendering, cursor-following patrol behavior, shape-aware dragging, and corner snapping. Users can converse, manage memory, or rename the pet via the right-click menu, and hide it using keyboard shortcuts. The plugin enhances perceived presence through visual feedback and is suitable for developers who need long-term context tracking or background task scheduling.
Read MoreAgent-Body: DeepSeek Harness's Organ-like Plugin Layer
Agent-Body is a plugin layer for DeepSeek Harness, designed to address context and routing overhead issues when agents have numerous tools. Through features such as architecture gating, deterministic routing, and failure attribution, it effectively manages tool architecture prompt tokens. The project supports local offline validation, is currently in early-stage development, and requires the DeepSeek Harness host to run.
Read MoreChatGPT Subscription Plugin: Use Your ChatGPT Plus/Pro Quota in DSH
DeepSeek Harness (DSH) offers a plugin called ChatGPT Subscription, designed to let DSH users directly access ChatGPT Plus/Pro models. It uses official OAuth login, supports selecting a model for conversations within DSH, and provides state management, multilingual support, and token synchronization. DSH must be restarted after installation for the plugin to take effect. Note that token writes may affect Codex CLI usage, and uninstalling does not delete authorization data.
Read MorefnOS CLI Skill: DeepSeek Harness Plugin for fnOS CLI
The fnOS CLI Skill is a DSH plugin for fnOS CLI, designed to provide developers with a bridge solution for managing the CLI through intelligent agents. This plugin installs the public bridge into the DSH environment and automatically resolves dependencies, enabling agents to execute fnOS CLI commands. To use it, manual authorization is required to obtain an access token, with attention to code sensitivity and TTY environment constraints. Installing the plugin alone does not automatically complete login; users must follow the specific authorization process for verification.
Read Moredsh-streamfold: Streaming Fold
dsh-streamfold is a display-enhancement plugin for the DeepSeek Harness client. It addresses the issue of the thinking process overwhelming the main text in long conversations by automatically expanding the latest thinking window, collapsing older windows into summaries, and interleaving them with the main text to achieve a frame-by-frame streaming collapse effect. The plugin runs entirely on the client side, is safe and non-invasive, and supports features such as customizable sparkle effects and pinned questions. It is ideal for users who frequently review the thinking process but do not want to be overwhelmed by long-form text.
Read MoreDeepSeek Harness Splash Screen Animation Plugin
The DeepSeek Harness plugin `deepseek-splash-animation` allows users to play a custom video or image when the application starts, automatically fading out after playback. It supports formats such as MP4, WebM, and GIF, and provides configuration options for muting, volume, fit mode, and a skip button. Users can install it via the command line and customize the content through the settings page or a configuration file. It is muted by default to comply with browser autoplay policies, and keyboard shortcuts such as ESC can be used to dismiss it.
Read MoreDeepSeek Harness Novel Forge Plugin
The `dsh-novel-forge` plugin, launched by DeepSeek Harness (DSH), provides engineering-grade hard constraints for long-form web novel writing through a “code enforcement over prompt compliance” design. It offers 20 `novel_*` tools and a right-side “Forge” panel. Core features include setting consistency management, a nine-stage workflow gate, AI-tone removal scanning, and platform review support. All revisions require proposal confirmation, and scanning diagnostics run entirely locally. After installation, restart the process to load the code constraints, ensuring writing standards and data security.
Read MoreDeepSeek Harness Plugin: dsh-background-by-model
`dsh-background-by-model` is a DeepSeek Harness plugin designed to automatically switch backgrounds based on the session model. It maintains a list of model rules and applies the corresponding background settings when a matching model name is detected, while also providing image carousel, global switching animation, and holiday background features. The plugin supports independent configuration for image layout, borders, and visual effects, with all data stored locally. Version v0.3.0 removed the video and background generation features. This plugin allows users to precisely manage the interface appearance to meet personalized needs.
Read MoreDeepSeek Harness Thinking Loop Guard Plugin: dsh-loop-guard
`mrweicodes/dsh-loop-guard` is a plugin that resolves the “thinking-only, no-action” infinite loop in DeepSeek Harness agents. Through a five-layer detection mechanism, it accurately identifies pure reasoning loops or repeated spamming, directly interrupts calls within the stream, and injects corrective prompts. The plugin does not interrupt the session, supports offline analysis, has a very low false positive rate, and requires DSH version 0.1.2 or higher. After installation, automatic resumption can be configured for unattended operation.
Read Moredsh-mcp-studio: MCP Services and Skills Manager for DeepSeek Harness
`dsh-mcp-studio` is an MCP server and Skills management plugin for DeepSeek Harness (DSH). It provides a unified extension panel that supports creating, reading, updating, and deleting MCP servers, importing and exporting configurations, and editing and searching Skills. The plugin works in both Web and Desktop environments and supports integrating the plugin marketplace when installing dshmarket. It requires Node >= 20 and can be added via the official command.
Read Moredsh-factory-provider: Integrate Factory Subscription Credits into DeepSeek Harness
`dsh-factory-provider` is an open-source plugin designed to integrate Factory (Droid) models into the native environment of DeepSeek Harness (DSH). It allows users to use Factory API keys directly for inference in DSH. The plugin supports multi-model grouping, upstream failover, secure local API key storage, token usage statistics, and subscription quota monitoring. Installation requires adding it via CLI and restarting DSH, with Node.js >= 20 and DSH version 0.2.0-rc.2 or later.
Read Moredsh-mcp: MCP Management Interface and Tool Search
dsh-mcp is a DeepSeek Harness client plugin designed to resolve tool-list bloat and cache invalidation caused by integrating multiple MCP servers. It provides visual management, fine-grained tool-level control, process-level environment variables, and OAuth authentication support. To save tokens, it offers on-demand retrieval (search) and full injection modes, and supports image pass-through and system certificate validation. Users can install it from the command line and manage servers under Settings → MCP, with support for declarative takeover, OAuth flows, and Tool Search mode.
Read MoreDSH Qoder Connect: Integrating Qoder’s Subscription Model
`masknull/dsh-qoder-connect` is a DeepSeek Harness (DSH) plugin designed to unify access to both the domestic and international product lines of Qoder. It features independent dual-variant configuration, real-time PAT validation, three-tier model catalog fallback, one-click context window switching, batch model toggle management, daily automatic check-ins, and sidebar display of quota and price multipliers. Installation requires DSH version 0.1.7-rc.1 or later, and users assume the risks associated with their own accounts.
Read Morebetter-dsh-session-deletetool: Truly Delete DeepSeek Harness Sessions
The DeepSeek Harness plugin **better-dsh-session-deletetool** addresses the limitation where the default behavior only archives sessions without deleting them, and provides true physical deletion. It supports single or bulk deletion through the session menu and workspace, and allows cascading deletion of sub-agents and derived conversations via a visual family-tree management view. The plugin can also clean up project directories and caches, and includes runtime checks, making it suitable for users who need to thoroughly reclaim disk space and manage complex session structures.
Read MoreDeepSeek Harness Plugin: CuiGenJi (Novel Writing Workbench)
The Cui Geng Ji plugin for DeepSeek Harness (DSH) is designed to address the difficulty AI faces in maintaining setting consistency in long-form creative writing. By providing a structured workspace, it supports manuscript editing, node-based plot planning, character and world-building management, and style presets, helping users collaborate with the AI to complete a work. It is compatible with the DSH Windows desktop version, supports one-click installation, and makes implicit materials—such as character motivations—explicit, enabling the AI to create based on complete profiles.
Read Moredsh-toolbox-ui: MCP and Skills Management Panel for DeepSeek Harness Web
The `dsh-toolbox-ui` client plugin for DeepSeek Harness (DSH) provides a floating panel that lets users manage MCP servers and local skills directly within conversations, supporting bidirectional operations between global and project-level configurations. The MCP management panel allows viewing status, starting or stopping connections, importing configurations, and managing credentials; the skill management panel supports viewing scopes, locating files, one-click toggling of the model auto-invocation switch, and running diagnostics. The plugin supports hot updates, enabling changes to take effect without restarting a session, making it suitable for developers who frequently adjust configurations.
Read Moredsh-workbuddy-connect: Connecting the WorkBuddy Model to DeepSeek Harness
The `dsh-workbuddy-connect` plugin allows users to directly use WorkBuddy models within the native DeepSeek Harness window without installing the desktop application. It features built-in login functionality, independent management for domestic and international versions, sidebar quota display, flexible model toggling, automatic check-ins, and log recording. Installation requires DeepSeek Harness core version `>= 0.1.7-rc.1`, with support for image input and manual reasoning tier detection (0.10.0-beta.5 or later is required for the TUI version). The plugin is maintained by masknull and licensed under the MIT license.
Read MoreJevDo: A Next-Generation Harness Architecture with Jev as the Scheduling Core
JevDo is a next-generation plugin architecture built on DeepSeek Harness, designed to address the problem of repeatedly reasoning about startup workflows during development. Its core separates scheduling logic from execution logic: **Jev** decides, based on history, whether to invoke a persisted **Action** (operation) or request the main model; **Action** saves reusable commands across sessions. Through the `jevaction_create` and `jevaction_validate` tools, the plugin turns the large model from the sole executor into one option among others, reducing the cost of invoking the main model. It is suitable for everyday bug fixing, CI error resolution, and release preparation, with later turns able to independently complete some tasks. The project is built with Node.js 24.14+ and requires DeepSeek.
Read MoreDeepSeek Harness QQ Bot Plugin Enhancements (dsh-qqbot)
`dsh-qqbot` is a QQ Bot plugin based on DeepSeek Harness that enables safe integration through the official API, reducing the risk of account bans. The plugin features core capabilities such as rich media sending and receiving, scheduled tasks, multi-instance personas, an affinity system, and a visual settings panel. It provides five extension points to support custom development. Installation must be performed in the `web` profile environment, making it suitable for deploying AI assistants in QQ groups and managing group activities.
Read MoreEAC-skin-loader: DSH UI Skin Loader
EAC-skin-loader is a standardized UI skin loader in the DSH ecosystem, designed to provide a zero-side-effect and recoverable theme-switching experience. The plugin includes 13 built-in skins and features core capabilities such as automatic discovery, globally exclusive switching, fault isolation, and recovery across restarts. It can be used via Git installation or local build and is visually manageable in DSH settings. The project follows a weak-constraint convention, supports 2 built-in example skins and 11 convention-based migrated skins, and places its licensing boundary at MIT for the engineering skeleton, with visual content distributed under its upstream licenses.
Read Moredsh-host-rsi: DeepSeek Harness Host-Side Memory and Self-Evolving Plugin
DSH-host-rsi is a host-side plugin designed to solve the “goldfish memory” and infinite-loop issues of local LLMs. It achieves model “self-evolution” through signal-ladder reliability weighting, prevents repeated mistakes via five-state delivery guardrails and a convergence-guidance mechanism, and offers persistent memory with crash self-healing. The plugin is non-invasive to model weights, supports project-level instruction deduplication, helps the model improve with use during sessions, and effectively saves computational resources.
Read Moredsh-jev-interceptor: Jev Risk Classification and Auto-Approval Plugin for DeepSeek Harness
The `dsh-jev-interceptor` plugin integrates non-generative Jev models into DeepSeek Harness (DSH), replacing the existing high-cost generative LLM classification approach. It delivers low-cost, high-efficiency tool-call decision support by evaluating risk, irreversibility, and injection suspicion with millisecond-level latency. Core features include risk classification, evidence-based automatic approval, session reference takeover, shadow/forced mode switching, and a fail-closed mechanism. Controlled through commands such as `/jev-on`, the plugin prioritizes safety and automatically delegates to DSH’s default behavior in the event of an error, making it suitable for scenarios that require strict risk control and faster response times.
Read Moredsh-effort-slider: DeepSeek Harness Inference Effort Slider
The `dsh-effort-slider` plugin transforms DSH's model inference level selector into a sci-fi-style slider with drag support. It adds an **Ultra tier** (locks the MAX level), a **Lightning mode** (the parent agent only handles orchestration, greatly increasing token consumption), and real-time display of child agents’ token output rate. The plugin also includes companion Skills and detailed technical documentation, and requires DSH Desktop 2.0.9 or later.
Read MoreIntroduction to dsh-packforge/dsh-pack-plugin
`dsh-pack-plugin` is a standalone npm plugin for DeepSeek Harness that provides integrated package management capabilities for the official desktop client through a bundle mechanism. With zero dependencies and zero source-code modifications, it offers complete export, installation, multi-profile switching, and isolation management features. At its core, it leverages Windows Junction technology to enable rapid environment switching and security validation, and it supports marketplace browsing and a task center, aiming to provide users with a standardized multi-environment management solution.
Read More