dsh.so
English

Install DeepSeek Harness

Install DeepSeek Harness (dsh) on your machine and start the Web UI. This guide follows the official README and was verified against it on 2026-08-13 — nothing here is guessed.

Prerequisites

  • Node.js ≥ 18 (LTS recommended). Check with:
zsh — dsh
$ node --version
v22.14.0
$

npm ships with Node.js. pnpm is optional but recommended for the ecosystem (npm install -g pnpm).

One-command install

npx @deepseek-ai/dsh web

That single command downloads the @deepseek-ai/dsh package (developer preview), boots the harness (Cordis plugin tree), and starts the Web UI. Per the official README, the Web UI is served at http://127.0.0.1:3080 by default.

Step-by-step

  1. Install Node.js if you don't have it (see Prerequisites).
  2. Run the command in a terminal: npx @deepseek-ai/dsh web
  3. Wait for boot. The harness composes plugins from the web profile. First run may take longer as packages are downloaded.
  4. Open the Web UI at http://127.0.0.1:3080 in your browser.
Expected output is illustrative — the command, the boot behaviour, and the default port are verified against the official README. Capture your own run and open a PR to replace this note with a real screenshot.

Verification

  • The Web UI loads at http://127.0.0.1:3080.
  • Inspect the actual plugin tree your machine boots:
zsh — dsh
$ dsh --profile web --dump-config
# every row printed can be replaced by a patch of your own
$

This prints the composed Cordis config — every row can be replaced by a patch of your own (see the Architecture guide).

Configure a model & run your first task

Per the official Web UI guide, once the server is running:

  1. Add a model — open Settings → Models, enter a DeepSeek API key, and save. The model route becomes usable immediately — no server restart.
  2. Choose a workspace — click Choose workspace and add the project directory where you started dsh. The session composer stays unavailable until a workspace is selected.
  3. Run a task — start a session and send something like “Summarize this repository and identify its main packages.” The agent can read and edit workspace files, run commands, delegate work, and maintain a plan; the Web UI asks before operations that require approval under the active permission policy.

The dsh process uses its invoking directory as the default filesystem location, so start it from the project you want to work on. Other providers and custom OpenAI-compatible endpoints are covered in the official model configuration guide.

Troubleshooting

Running into something not covered here? See the full Troubleshooting guide — port conflicts, profile errors, Windows quirks, and config debugging.

  • Port already in use — another process holds 127.0.0.1:3080. Find it and stop it, or check the CLI reference for a port flag.
  • Stale npx cache — npx @deepseek-ai/dsh web caches packages; if behavior looks outdated, clear the npx cache or run from a fresh source checkout.
  • Preview instability — dsh is in developer preview and iterating rapidly; pin the version you rely on and track the Changelog.

Next steps