Install Guide

How to Install DeepSeek Harness

Everything you need to run the official open-source harness on macOS, Windows or Linux — from zero to your first agent task.

Heads up: the official way to run DeepSeek Harness today is a command-line tool that assumes Node.js and a terminal. If that sounds like friction, you are exactly who our planned one-click Desktop app is for — join the waitlist while you try the steps below.

Before you start

You will need three things:

Node.js installed (the project is installed and run through npm/npx).
A terminal on macOS (Terminal or iTerm), Windows (PowerShell) or Linux (any shell).
A model API key — a DeepSeek API key works out of the box; other providers and OpenAI-compatible endpoints are supported too.

The harness itself is free and open source (MIT license). The only cost is model tokens billed by the provider you configure with your own key.

Option A — Quick start with npx

Step 1

Run the one-liner

From any directory, run:

npx @deepseek-ai/dsh web

npx downloads the package on first run, then starts the Web UI. The command prints the URL — by default http://127.0.0.1:3080. Open it in your browser.

Option B — Install from source

Prefer to run a repository checkout, or want to poke at the code? This path needs pnpm:

Step 1

Clone the repository

git clone https://github.com/deepseek-ai/deepseek-harness.git
cd deepseek-harness

Step 2

Install and build

pnpm install
pnpm run build

Step 3

Start the Web UI

pnpm dsh web

Configure a model

With the Web UI open, the next steps live in the UI itself:

Open Settings → Models.
For DeepSeek: enter your API key in the DeepSeek card and save. For another provider, choose Add provider and pick e.g. Anthropic or OpenAI.
For a company gateway or self-hosted server: choose Add a custom provider and supply a provider ID, base URL, API protocol, credential and at least one model. It must be OpenAI-compatible or use one of the supported protocols.

Keys are write-only — the UI shows a redacted descriptor after saving, and the key is stored in $DSH_HOME/.credentials.yaml. The model route becomes usable immediately; no server restart needed.

Choose a workspace

Click Choose workspace, add the project directory where you started dsh, and select it. The session composer stays unavailable until a workspace is selected, because the agent needs to know where it may read and edit files.

Run your first task

Start a session and send a prompt such as:

Summarize this repository and identify its main packages.

The agent can read and edit workspace files, run commands, delegate work and maintain a plan. When the active permission policy flags an operation, the UI asks for your approval before the agent proceeds.

You are running. That is the whole flow: run dsh web, add a key, pick a workspace, start a session. Everything after that is in the UI.

Troubleshooting

Nothing prints a URL? Check that the command completed without errors, and try http://127.0.0.1:3080 directly.
Port already in use? Stop whatever is occupying 3080, or run the server from a different working directory/configuration.
Model requests fail? Confirm the key was saved under Settings → Models, and that the provider you selected matches the key you entered.
Windows users: use PowerShell and make sure Node.js is on your PATH; the npx one-liner behaves the same as on macOS/Linux.

What about the Desktop app?

This site exists because we believe the installation, setup and update steps above are exactly the friction that keeps great tools away from most people. The planned DeepSeek Harness Desktop app wraps the same open-source engine with one-click installs, guided provider setup and automatic updates — your key stays local, and the engine underneath stays the same. Join the waitlist to reserve early access, and read the tutorial to understand what the harness can do while you wait.

Tutorial → Core concepts: plugins, presets, session logs and the Web UI. Back to Home → The waitlist, pricing hypothesis and FAQ.