webdsh

webdsh

DeepSeek Harness in a browser tab — the real agent, real Node, no server to run.

DeepSeek Harness (dsh) is an agent harness where everything is a plugin. dsh web runs a Node host and serves a browser client to it. webdsh is that, as static files — the host runs inside the page, and the agent's commands run in WebContainers: Node itself, in the tab.

• ⚡ Nothing to run. No server, no install, no local Node — the harness boots in the page.
• 🖥️ Real Node, real Python. npm install and pip install both work, and the terminal and the agent share one container.
• 💾 Or a whole PC. Switch the runtime to v86 and the session runs FreeDOS, Windows 1.01, Windows 3.1, Windows 98 or Linux — emulated x86, booted from a disk image, with the tool set that machine actually has.
• 🧩 Real plugins. Install from npm, a tarball, GitHub, or a path — from the browser.
• 📦 Real dsh. The published @deepseek-ai/* packages, unmodified: 114 of 129 rows compose exactly as dsh web composes them.
• 🔒 Yours. Files, sessions and keys live in your browser's storage. Nothing is uploaded.

Table of Contents

• Background
• Install
• Usage
• Maintainers
• Contributing
• License

Background

Nothing here is a fork of dsh. The agent loop, tool registry, model adapters and the entire web client come from npm at install time; the only modification is a cordis.patch.yml layer — the mechanism dsh documents for exactly this.

What this repository adds is the platform underneath: a synchronous POSIX filesystem mirrored to IndexedDB (src/vfs), node:* implemented over it (src/node), the two runtimes a session can run on — WebContainers and an emulated x86 PC (src/runtime) — an in-page virtual server for /api plus the CORS policy every outbound request goes through (src/net), and the plugins this build ships (packages/).

Six composition rows are swapped, each because the shipped one names something a page cannot have — or, in the shell's case, cannot honestly describe. Four more are reconfigured rather than replaced, including the one that decides whether this deployment can open a path at all. npx tsx scripts/alignment.ts prints the whole difference.

Install

Nothing to install — open the page. To run it yourself:

npm ci
npm run build        # → dist/
node scripts/serve.mjs 4173

Node 22 or newer. dist/ is plain static files with relative URLs, so it works at a domain root, a project path, or a local directory.

Usage

Open the page, choose a workspace, start talking.

• Files — the sidebar action above the terminal. Browse the workspace, open a file, drop files in, and take things out: one file as itself, a directory or a tick-box selection as a zip. It is the same filesystem the agent and the terminal use, not a copy. A file path the assistant names in the chat opens here when you click it.
• Terminal — Ctrl+` or the sidebar action. It is Node in this tab, and the same machine the agent's tools run in. Closing it hides it; the session, its scrollback and its working directory are still there when you reopen.
• Python — python3 and pip are there for both of them. The first call fetches a 14 MB interpreter; after that it is stored, and packages installed with pip survive a reload. Write python3: jsh aliases python to it and loses the quoting on the way, so python -c "…" is a syntax error.
• Runtime — the sidebar action below the terminal. A session runs on one machine and this is where you pick it. The default is the Node container above; the alternative is an emulated 32-bit PC, and the panel is that PC's screen — live, with a working keyboard, showing what the assistant is doing. Five machines need no setup at all: Linux (busybox on a serial console, the shortest way here to a real POSIX shell), FreeDOS, MS-DOS 7, Windows 1.01 and KolibriOS. Eleven more boot exactly the same way but are not this deployment's to serve a disk for — Windows 2.03, 3.0, 3.1, 95, 98, ME, NT 4.0 and 2000, MS-DOS 6.22, Buildroot Linux, and Arch Linux, which is a 2022 kernel with bash, python3 and gcc and resumes from a saved machine in about two seconds. For those, open a disk image from your computer or point the panel at a host that has one. Nothing is downloaded until you choose a machine, and a change applies on the next load.
• Plugins — Settings → Plugins, or /plugin add <package> in the composer. Takes an npm name, a tarball URL, owner/repo#ref, or a path. The Installed tab turns one off or removes it; the composition is fixed at boot, so a change applies on the next reload and the panel says so.
• Models — 42 models across six routes are registered up front and need no account, so the page answers before it asks for anything. Settings → Models offers the rest of the provider catalog; typing a key is the whole of configuring one.
• Network — Settings → Network picks the CORS proxy, used only after a direct request has actually failed, and reported to you when it is.
• Persistence — workspace, sessions and transcripts survive a reload. window.dsh.exportFs() downloads a zip; window.dsh.reset() clears it all.

On an emulated machine the assistant is given different tools, because it is a different machine: sh on the Linux guest, dos on the DOS ones — both reading a real character stream, so a command's whole output comes back rather than the last 25 lines of it — and vm_screenshot, vm_screen, vm_key, vm_type, vm_mouse and vm_wait everywhere, which is the whole of the tool set on a guest that only draws pixels. There is no jsh, no Node and no Python in that session, and the guest's disk shares nothing with the workspace your file tools read. The panel and the tools say so.

Worth knowing: the container's shell is jsh, not bash, and it ships no git; python3 is CPython 3.14 compiled to WebAssembly, fetched on first use and kept afterwards, so it has pip but no compiler, no subprocesses and no sockets; and a host that refuses browsers is reachable only through the proxy, which does not extend to the container.

Maintainers

@futrime

Contributing

Issues and PRs welcome at futrime/webdsh. Commits follow Conventional Commits; npx tsc --noEmit and npx tsx scripts/e2e.ts should pass first.

License

Apache-2.0 © Zijian Zhang. The @deepseek-ai/* packages it composes are published by DeepSeek AI under their own terms.