A native dsh plugin that wires nine IM bots (Feishu, WeChat, Slack, Telegram, Discord, WhatsApp, …) and an AI Office Connector into your DeepSeek Harness session.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-22
License
MIT
01What can it help you accomplish?
Let people chat with your DeepSeek Harness from everyday IM apps
Connect nine IM channels (Feishu/Lark, WeChat, DingTalk, WeCom, QQ, Slack, Telegram, Discord, WhatsApp) so messages in those apps drive a dsh session with streaming thinking / tool progress / replies, plus image understanding (JPEG, PNG, WebP, ≤5 MB each, ≤20 MB per message)
Teams that want to operate a dsh agent from familiar IM clients instead of the web UI, with each bot keeping its own workspace and session
Connect a local dsh to a public AI Office without exposing your network
The AI Office Connector makes the local Harness dial out to a public Office over SSE — no public IP, port-forward or WebSocket server needed — and the Office only ever receives aliases, never your local absolute paths
Users who want remote / office access to a home or LAN dsh without opening inbound ports
02How to install into DeepSeek Harness
Prerequisites
- DeepSeek Harness with the web interface (`dsh web`) and a web profile
- For each channel, the bot's QR scan or credentials (e.g. Feishu App ID + Secret, Telegram BotFather token, Discord / Slack bot token, WhatsApp device scan)
Installation steps
- 01
Add the plugin to your web profile: `dsh plugin --profile web add -w @xmanrui/dsh-im`
$ dsh plugin --profile web add -w @xmanrui/dsh-im
- 02
Restart `dsh web`, then open Settings → Plugins → IM Bot
$ dsh web
- 03
On each channel's page, complete the built-in QR scan or credential setup following the in-page guidance
Verify the integration
- After restart, open Settings → Plugins → IM Bot and confirm the nine channels and the AI Office Connector entry are visible; a connected bot answers `/status` when it can reach DeepSeek Harness
03DSH integration and capability boundaries
Native DeepSeek Harness (`dsh`) web plugin `@xmanrui/dsh-im`. One plugin and one settings entry connect nine IM channels (Feishu / Lark, WeChat, DingTalk, WeCom, QQ, Slack, Telegram, Discord, WhatsApp) and the AI Office Connector to your local dsh, so messages from those apps drive a dsh session with streaming thinking, tool progress and replies, plus image understanding.
Nine IM channels (multi-bot)
text and images sent to a bot in Feishu / Lark, WeChat, DingTalk, WeCom, QQ, Slack, Telegram, Discord or WhatsApp→the message drives a dsh session with streaming thinking, tool-progress and reply shown in-platform; each bot keeps independent credentials, connection state, workspace and session binding
creates and binds dsh sessions per bot / chatAI Office Connector
a public AI Office Base URL and a generated Device Token→the local Harness dials out over a heartbeat + SSE stream (protocol `office-harness.v1`), pulls jobs, runs a leased dsh session and streams status / tool names / incremental text back; tool approvals round-trip via SSE
opens an outbound connection to the Office; Device Token is stored only in the Host credential storeBot command & session control
slash commands typed in the IM chat (`/new`, `/status`, `/models`, `/model`, `/preset`, `/stop`, `/steer`, `/compact`, `/workspace`, `/session`, …)→manage the bound dsh session, switch models and Agent Presets, steer or stop a running task, and switch workspace — without leaving the chat app
may create / rebind dsh sessions and change the bot's workspace or presetImage understanding
JPEG, PNG, WebP and GIF-sent-as-image with an optional caption→images are forwarded to the dsh session for the model to read; single image ≤5 MB, total per message ≤20 MB
04Who is it for? When not to use it?
Good for
- Teams that want to operate a dsh agent from familiar IM clients instead of the web UI, with each bot keeping its own workspace and session
- Users who want remote / office access to a home or LAN dsh without opening inbound ports
Not for
- All Secrets and Tokens are submitted only to the local Harness Host and written to a protected credential store; status and bot-list responses never return them. The IM management RPC accepts only the loopback browser by default. The `trusted-host` LAN mode lets anyone on that network view bot status, scan QR, submit credentials, reconnect or delete bots — enable it only on trusted networks.
- The AI Office Connector only sends aliases (device ID, Office Origin, workspace alias, instruction-preset alias) to the Office; the Office never receives local absolute paths, so it cannot read your local filesystem. Workspace / session lists come from the global Harness Host and may include other bots' or projects' local paths and sensitive metadata — only expose bots to trusted users.
05Compatibility, maintenance and safety notes
- Requires the DeepSeek Harness web interface (`dsh web`) and a web profile. Each channel needs its own bot (QR scan or credentials), and Slack's desktop client may intercept leading-slash messages — prefix the command with a space (e.g. ` /presetlist`).
- All Secrets and Tokens are submitted only to the local Harness Host and written to a protected credential store; status and bot-list responses never return them. The IM management RPC accepts only the loopback browser by default. The `trusted-host` LAN mode lets anyone on that network view bot status, scan QR, submit credentials, reconnect or delete bots — enable it only on trusted networks.
- The AI Office Connector only sends aliases (device ID, Office Origin, workspace alias, instruction-preset alias) to the Office; the Office never receives local absolute paths, so it cannot read your local filesystem. Workspace / session lists come from the global Harness Host and may include other bots' or projects' local paths and sensitive metadata — only expose bots to trusted users.
MIT · actively maintained (latest release v1.0.2, 2026-08-22)
06Frequently asked questions
How do I install dsh-im?
Run `dsh plugin --profile web add -w @xmanrui/dsh-im` in your web profile, restart `dsh web`, then open Settings → Plugins → IM Bot. For each channel, follow the in-page QR scan or credential setup. (An unstable GitHub-source installer `npx -y github:xmanrui/dsh-im install` also exists, but the npm stable build is recommended.)
Which IM channels are supported?
Nine: Feishu / Lark, WeChat, DingTalk, WeCom (Enterprise WeChat), QQ, Slack, Telegram, Discord and WhatsApp. Each channel can connect multiple bots, and every bot keeps its own credentials, connection state, workspace and session binding.
How does an IM message reach DeepSeek Harness?
After connecting a bot, messages in that chat drive a dsh session. The plugin shows streaming thinking, tool progress and replies in-platform. You manage the session with slash commands: `/new` (new session), `/status` (connection check), `/models` & `/model` (switch model), `/preset` (Agent Preset), `/stop`, `/steer`, `/compact`, `/workspace`, `/session`, and more.
What is the AI Office Connector?
It lets your local Harness dial out to a public AI Office over a heartbeat + SSE stream (protocol `office-harness.v1`), so you get remote / office access without a public IP, port-forward or WebSocket server. The Office only ever receives aliases — never your local absolute paths.
Is it safe to give the bots my credentials?
Secrets and tokens are submitted only to your local Harness Host and written to a protected credential store; status and bot-list responses never return them, and the IM management RPC accepts only the loopback browser by default. Note: anyone who can message a bot can run session / workspace / preset commands, so restrict bot visibility to trusted users, and only enable the `trusted-host` LAN mode on trusted networks.
07Related DSH workflows
awesome-gpt-image-2
by freestylefly
Prompt-as-Code engine and template library for GPT-Image2, with 470+ reverse-engineered cases, 20+ production templates and distilled agent skills.
voyager
by nagi-ovo
Browser extension that enhances Gemini, AI Studio, Claude and ChatGPT with a prompt manager that works across any web UI, DeepSeek Harness included.
wegent
by wecode-ai
Plan, build, and deliver with an open-source, self-hostable AI workspace for coding, collaboration, and automation.
chatgpt-gemini-timeline
by houyanchao
Chrome extension that adds folders, a prompt library, and a conversation timeline to ChatGPT & Gemini chat pages — works alongside DeepSeek Harness.
08Data and sources
通过扫码、App Manifest 或已有机器人凭据把 IM 机器人接入 DeepSeek Harness
dsh plugin --profile web add -w @xmanrui/dsh-im
一个插件、一个设置入口，统一管理九种 IM 渠道和 AI Office Connector。
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-22. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.