Plugins

Adds an ego-webqa MCP server that audits deployed pages across the mobile viewport matrix and returns structured repair evidence; needs uv/uvx on PATH and a Playwright Chromium headless runtime on the same machine.

Windows only (binds user32.dll via koffi; inert on macOS/Linux). Remembers the DeepSeek Harness main window's size AND position: restores both on every launch (re-asserted during a grace period to beat the default layout), saves stable manual moves/resizes as the new lock, never records minimized/maximized/fullscreen states, never adopts the default rect DSH itself places the window at (so the app's own layout resets can never poison the lock), and stays on-screen when monitors change. Compatible with DSH 0.2.1-rc.3+.

Starts a BaseMind MCP server for each session's workspace and registers the code-intelligence tools it advertises, with no MCP row to configure. Requires BaseMind installed and on PATH (basemind serve); without it no tools are registered.

Windows x64 desktop tools for DeepSeek Harness to restore and foreground existing windows, observe screenshots and UI Automation elements, left-click, and type text or supported keys, with repeatable control sessions and physical Esc to stop.

Unity-MCP bridge for DSH: dormant from boot, /unity-mcp-tools connects the Unity editor MCP server (com.ivanmurzak.unity.mcp) and registers the tools currently enabled on the Unity side; ships a per-plugin-version command catalog plus /unity-mcp-install and /unity-mcp-check, and re-fetches tools/list itself after a tool-set change because DSH sessions never receive list_changed.

LAN phone drop: a QR code on the computer screen opens a page on the phone, and photos or scanned documents land in a computer folder over the local network, with no tunnel and no cloud in between. Document mode binarizes locally in the browser (adaptive threshold plus speckle removal and edge crop) so an ordinary phone photo comes out as a black-and-white scan. Adds a "phone drop" entry at the foot of the DSH sidebar that opens the QR page, plus a phone_drop tool for the assistant; the same CLI also runs standalone without DSH.

Adds a calculate tool backed by a hand-written parser with no eval, batch evaluation and an exact-decimal mode for money math.

Links the current DeepSeek account in DSH to a GitHub identity through the local GitHub CLI, a personal access token, or OAuth device flow, then exposes a scoped github tool for listing repositories and reading, creating, or commenting on issues.

Five DSH utilities in one plugin: drive existing sessions over QQ (needs an OneBot 11 implementation such as NapCat running alongside), ask another session a question with a context guard, pause the running turn, restart DSH on Windows with optional auto-resume (via the install-restart.ps1 scheduled task), and inject peak-hour token discipline.

Team shared knowledge base: one member runs a zero-dependency Hub, everyone else installs this plugin and joins with one invite code. Uploads are filed into ten categories with a summary, confidence and points; 15 agent tools let members upload, search, recall, comment and read them. Important items are pinned with a hand-written recommendation until the member has read them, and each account also gets a private local library, collections, a weekly digest and version diffs.

Rejects foreground waiting in DSH agents: waiting commands (sleep, wait, tail -f, watch, ping without -c, infinite loops), blocking wait tools (wait_agent, job_output with wait: true), progress polling (the same check command repeated within a window) and commands that need human input/confirmation (sudo without -n, ssh without BatchMode, login flows, interactive editors, docker -it) are refused with an actionable retry to run in the background or end the turn. Configurable budget and polling window, message language follows the system (zh/en), zero runtime dependencies.

Alternative front-end for the upstream REA (Reverse Engineer Anything) MCP server: instead of registering its 207 tools (about 775,000 bytes of tool schemas in every request), it exposes three calls — rea_tools lists the catalogue grouped by domain, rea_help returns one tool's schema, rea_call passes arguments through unchanged — and spills oversized results to files instead of truncating silently. Ships no engines: REA (rea-agents 6.x) and Ghidra 12.1.x must be installed separately, and the analysis itself is REA/Ghidra's, not the plugin's.

Local document library for DSH Desktop (Windows only, installed with the bundled PowerShell script): upload files once, then every conversation can search and read them through four tools (library_list / library_search / library_read / library_save); upload, download and batch download live in the Settings panel. Files stay on the machine unless you enable the optional vision model, which sends images to the endpoint you configure.

Exposes Windows display geometry and the foreground window rect in physical pixels.

Connect DSH to a local BiliMemory backend for Bilibili search, collection, status and login tasks, A-Memorix shared memory health checks, plus QQ link status, QQ user allowlists and QQ watcher restart. The repository includes a Windows one-click installer that installs or repairs MaiBot A-Memorix when missing, then configures local-bge-small-zh, embedding and shared memory. It also provides install-memory.cmd for memory-only repair. Requires Windows, DSH Desktop 0.2.1-rc.2+, Node.js 22.19+, Python 3.12, Playwright/Edge, FFmpeg, a running BiliMemory API at 127.0.0.1:8765, a local embedding service at 127.0.0.1:11435, a Bilibili login, and MaiBot/NapCat for QQ. Voice, RVC and cover features are excluded.

Registers a `dispatch_task` tool that hands scoped or routine work to an isolated subagent, waits for it to finish, then checks the expected files and runs an optional verification command before returning a short report to the main conversation.

Reads the local NTQQ (QQ 9.9.x) message database on Windows and Linux and exports group and private chat history to txt, csv, json, md, html or sqlite. Ships eight model tools (qqnt_status, qqnt_extract_key, qqnt_list_groups, qqnt_list_chats, qqnt_search, qqnt_export_group, qqnt_export_chat, qqnt_export_batch) and a local web UI on 127.0.0.1:46728 with an incremental SQLite index, statistics, ZIP batch export, sender redaction, media bundling, per-day splitting, export presets and AES-256 password ZIPs. Needs python3 with sqlcipher3 (the settings page can create a private virtualenv), and the database key is read from the running, logged-in QQ client.

Moss text-to-image inside DSH — describe the picture in chat and get it back; needs a Moss API key from platform.moss.cn, and generation is billed to your own account.

Siemens NX integration for dsh on Windows (requires a local Siemens NX install; tested only with NX 2512, and the FE tools need Simulation/Nastran licences): 71 tools that build and edit solids, surfaces, sketches, holes and tubes, run NX Nastran statics and modal studies, roll up assembly BOMs with pairwise clearance checks, and print A3 sheets with a real GB frame to PDF, all in a background NX session driven by JSON commands.

Connects AutoCAD, ZWCAD, GstarCAD and BricsCAD to DeepSeek Harness over MCP, bundling the CAD server and building a private Python environment on first run. Requires Windows, one of those CAD applications installed, Python 3.10+ on PATH, and CAD running as the same Windows user and at the same privilege level as DSH.

Runs the local nanobot CLI agent as a DSH subagent through a nanobot_run tool: each call carries one self-contained task and returns only the agent's final answer, either as a fresh process or through a persistent OpenAI-compatible server.

Turns archived sessions into durable memory instead of dead disk. Distils each archive into reusable skills, a fact ledger with ADD/UPDATE/NOOP decisions plus validUntil and supersededBy so stale facts retire instead of piling up, and topic clusters by single-linkage similarity. Also digests sessions that are still in progress through a read-only seq watermark, leaving the session log byte-for-byte untouched, and can self-check the growing memory to merge duplicate facts, retire stale ones and consolidate duplicate skills — measuring only unless you opt in, and never hard-deleting, only superseding.

Memory the assistant writes and maintains itself — diary, index, self-description, the live emotion value — put back in front of it at the start of every session. It reads only the sections you name, never summarises, fills in missing starter files, and stamps a start date into an undated anchor in place — apart from that one additive edit it never changes an existing file. The writing side ships as skills.

Trim tools for long chats: trim_context / estimate_context_tokens keep the system persona and the last message, drop the oldest user/assistant pairs to fit a turn or token budget, and report before/after estimates; the conversation is not trimmed unless something calls the tools.