dsh web is the DeepSeek Harness web UI, a locally running web app. After the service starts, open http://127.0.0.1:3080 in the browser — it is the main interface for dsh, hosting settings, chats, the Plugin Center and logs in one place.

How do I open dsh web?

Three ways: run npx @deepseek-ai/dsh web without a global install; run dsh web after npm install -g @deepseek-ai/dsh; or, if the service is already running, type http://127.0.0.1:3080 in any browser. Keep the terminal window open — closing it stops the service.

What are the main areas of the dsh web interface?

Four core areas: Settings (model providers and API keys), the chat window (conversations and history), the Plugin Center (install and manage plugins, appears after installing DSH Plugin Hub and restarting), and system logs plus diagnostics for troubleshooting.

How do I configure a model in dsh web?

Open Settings → model config, pick a provider, fill in the endpoint and key (for example an OpenAI-compatible API), save, and the model becomes available in the chat window. Per-provider details are in the official docs.

How do I install a plugin in dsh web?

First add the market entry from the command line with dsh plugin --profile web add dsh-plugin, restart dsh web, then go to Settings → Plugin Center → search → Install → confirm. Alternatively install any plugin directly with dsh plugin --profile web add .

How to use dsh web: open it, tour the UI, install plugins

dsh web is the DeepSeek Harness web UI: start the service and open http://127.0.0.1:3080 in the browser. This guide covers three ways to open it, a tour of the main areas (settings, chat, Plugin Center, logs), and the common workflows for configuring a model and installing plugins — closing with a troubleshooting table.

Overview: what dsh web is

dsh web is a locally running web app — no desktop client to install, just a browser. Once started, everything happens in one interface (source). The four core areas at a glance:

Area	What it does	When to use it
Settings	model, provider and key config	first run, switching models
Chat window	talk to models, manage history	daily use
Plugin Center	browse, install, manage plugins	installing plugins
Logs and diagnostics	review activity, troubleshoot	when something fails

Three ways to open dsh web

Start by command, keep the service running, or visit directly — pick whichever fits. Three ways:

Way 1: start with npx (no install)

Prerequisite: Node.js on the machine; node -v prints a version;
Open a terminal and run:

bash

npx @deepseek-ai/dsh web

The first run downloads the package and initializes the web profile; wait for the "service started" message;
The browser opens http://127.0.0.1:3080 automatically; finish the first-time setup.

Way 2: start after a global install

Prerequisite: npm install -g @deepseek-ai/dsh has been run before;
Run dsh web in the terminal;
The browser opens http://127.0.0.1:3080.

Way 3: visit directly when the service is already running

Prerequisite: the dsh web service is running in some terminal (started by either way);
Type http://127.0.0.1:3080 in any browser address bar and press Enter.

Fallback: dsh: command not found means there is no global install — switch to Way 1. If the address won't open, confirm the terminal service is still running (closing the terminal window stops the service).

Tour of the main areas: settings, chat, Plugin Center, logs

Think of the interface as configure → use → extend → troubleshoot. Tour:

Settings: configure model providers and API keys — the first stop on a fresh install; port and proxy options live here too;
Chat window: talk to models and manage session history — the daily workspace;
Plugin Center: appears after installing DSH Plugin Hub and restarting; browse, install, update and uninstall plugins from here;
System logs and diagnostics: review install/update/uninstall trails and run diagnostics for troubleshooting.

Common workflows: configure a model, install plugins, read logs

Three high-frequency operations, learned once. Steps:

Workflow 1: configure a model (change settings)

Open Settings → model config;
Choose a provider, fill in the endpoint and key (e.g. an OpenAI-compatible API);
Click Save; the interface confirms;
Return to the chat window, pick the model you just configured and start talking.

Workflow 2: install a plugin

Add the market entry from the command line: dsh plugin --profile web add dsh-plugin;
Restart dsh web and refresh the page;
Open Settings → Plugin Center and search or browse to the target plugin;
Click Install on its card, review the plugin name and source in the confirmation dialog, then confirm;
When the live progress finishes, refresh the page — the plugin is active.

Workflow 3: read the logs

Open Settings → System logs;
Filter by category (install, update, uninstall, runtime) and level;
Locate the relevant record and copy the full error text for troubleshooting.

Troubleshooting common issues

Can't open it, changes don't apply, errors thrown — locate the problem in the table first. Common issues:

Issue	Likely cause	Fix
Port 3080 won't open in the browser	service not running or already exited	go back to the terminal and confirm the service is up; closing the window stops it
`dsh: command not found`	no global install	use `npx @deepseek-ai/dsh web`, or run `npm install -g @deepseek-ai/dsh` first
Plugin Center missing from the UI	DSH Plugin Hub not installed, or not restarted	`dsh plugin --profile web add dsh-plugin`, then restart dsh web
Plugin not active after install	page not refreshed, or wrong profile	refresh the page; make sure the command used `--profile web`
Chat messages error out	model not configured, or wrong key	check the endpoint and key in Settings → model config, save, retry
Setting changes don't apply	not saved, or a restart is required	click Save; some settings only take effect after restarting dsh web

Quick recap

In short: the web UI is the gateway — configure models in Settings, install plugins in the Plugin Center, check logs when something fails, and confirm the service is running if the page won't open. Four reminders:

Configure a model first — without a provider configured, the chat window cannot work;
Restart and refresh after installing plugins — restart dsh web and refresh after any install or uninstall to make sure it takes effect;
Logs keep records — every install, update and uninstall is written to the system logs; read the logs first when something goes wrong;
Keep the terminal open — the dsh web service runs in the terminal; closing the window stops it.

Sources: DeepSeek Harness docs - Quickstart, dsh CLI README, dshplugin/dsh-plugin-hub