DeepSeek Harness
How to Install and Run DeepSeek Harness
Follow the verified npm and source-checkout paths for DeepSeek Harness, then configure a model and workspace in the Web UI.
What you will learn
- Use npx for a quick Web UI trial; use pnpm when developing from source.
- A first Web UI task requires both a saved model key and a selected workspace.
- Node and pnpm versions are part of the reproducible setup.
Before you start
- Basic HTTP and API knowledge
Leave with a concrete implementation checklist and a testable starting point.
Key takeaways
- Use npx for a quick Web UI trial; use pnpm when developing from source.
- A first Web UI task requires both a saved model key and a selected workspace.
- Node and pnpm versions are part of the reproducible setup.
Pick an installation path
For a quick local trial, install Node.js and run `npx @deepseek-ai/dsh web`. The README says this starts the Web UI at `http://127.0.0.1:3080` by default and opens a browser for a local launch; add `--no-open` when the process should only serve.
For source work, clone the repository, run `pnpm install`, build with `pnpm run build`, then launch with `pnpm dsh web`. The final command uses the built artifacts without rebuilding, which makes the build step an explicit checkpoint.
First run in the Web UI
After the server starts, open Settings → Models, enter a DeepSeek API key, and save it. Then choose the workspace directory where you started dsh; the guide says the session composer remains unavailable until a workspace is selected.
Start a session with a concrete task such as asking the agent to summarize the repository and identify its main packages. The UI can read and edit files, run commands, delegate work, and maintain a plan, while approval prompts depend on the active permission policy.
Troubleshoot with evidence
If the UI does not open, use the printed host URL and check whether port 3080 is already occupied. If a source checkout fails before the app starts, verify the Node engine and the pinned pnpm version in the development guide before changing dependencies.
Do not paste real credentials into issues or commits. The development guide documents `DEEPSEEK_API_KEY` and optional `DEEPSEEK_BASE_URL` environment variables for demos and tests, and says real-API suites skip when the key is absent.
Decision guide
| Criterion | Option A | Option B |
|---|---|---|
| Use this path | npx when evaluating the current published package | pnpm source flow when changing code or docs |
| Main checkpoint | Server prints a local URL | typecheck/build succeeds before dsh starts |
Implementation steps
- 1
Install a supported Node.js version and Corepack-enabled pnpm for source work.
- 2
Run the npm command or clone, install, build, and launch from source.
- 3
Configure the DeepSeek key under Settings → Models.
- 4
Choose a workspace and run a small read-only task before enabling edits.
Copy-ready example
# Fast local launch
npx @deepseek-ai/dsh web --no-open
# Source checkout
git clone https://github.com/deepseek-ai/deepseek-harness.git
cd deepseek-harness
pnpm install
pnpm run build
pnpm dsh webFrequently asked questions
Which port does the Web UI use?
The README documents http://127.0.0.1:3080 as the default local address. Use the URL printed by the process if your launch context differs.
Do I need an API key to install it?
No for installing the package, but the Web, headless, and ACP demos need a DeepSeek API key to call the real API.
Sources
- DeepSeek Harness README run guideSource checked 2026-08-27
- DeepSeek Harness development guideSource checked 2026-08-27
- DeepSeek Harness Web UI guideSource checked 2026-08-27