DSH Marketplace

dsh-browser

Lum1104/dsh-browser

Chrome sidebar extension that lets DSH operate your browser directly, no vision capabilities required.

84269TypeScriptMITSource

Install

Add dsh-browser to DeepSeek Harness

Install into

via GitHub · GitHub source

Installing from GitHub runs the project's build script, which pnpm blocks until you allowlist it — run the command once and pnpm prints the exact key to add under `allowBuilds` in ~/.dsh/profiles/web/pnpm-workspace.yaml.

What happened when we ran it

Installed cleanly when we ran it

Every command here is run in a throwaway container against a clean profile, and the result is whatever the harness recorded — not a guess from the source. Last run 11d agoon DeepSeek Harness 0.2.2-rc.1.

Show it in your README

install verified — dshmarketplace

For maintainers: the badge serves this listing's latest sandbox verdict, so a re-run updates it on its own — and it links readers to the full result here.

Due diligence

Before you install dsh-browser

  • Source of record: Lum1104/dsh-browser — present in the community registry that DSH's own plugin market installs from.
  • Licensed under MIT.

Security

What installing dsh-browser runs

Build scripts

pnpm skipped these dependency build scripts, and the plugin registered without them:

  • @deepseek-ai/dsh-subprocess-local@0.2.0-rc.2

Under pnpm 11 the install stops and asks you to approve them instead; an approved script runs with your user account's permissions.

Source: our sandbox install on DeepSeek Harness 0.2.2-rc.1, 11d ago.

Install hooks

package.json declares no preinstall, install or postinstall script.

Source: package.json on GitHub, read at our last sync. prepare is not recorded; it runs only on a GitHub install, and shows up under build scripts when pnpm blocks it.

README mentions

shell commands (bash, exec, spawn …); API keys, tokens or passwords

Source: a keyword match on the README — a hint about what the plugin touches, not an analysis of its code.

Not an audit. These are the record of one install and two text scans; nobody has read this plugin's code on your behalf. Plugins run with your agent's permissions.

The AI take

What it is — Provides a sidebar for DSH to directly operate your existing browser tabs.

Who it is for — Suitable when you need to control browser tabs using DeepSeek Harness. It is not suitable if you use a headless browser environment.

Watch out — Sandbox test passed: installed in a fresh profile and registered by harness into the profile. It will execute shell commands and requires providing an API key or token. No obvious issues found.

The verdict — I would install it because it completed all runs in the benchmark with lower latency than Playwright.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.

What dsh-browser does

Structured page data flows through an authenticated bridge into numbered controls and returns as browser actions.

dsh-browser is a DeepSeek Harness plugin that connects DSH to a Chrome tab you already use, allowing the agent to read pages and operate their controls without visual input. It combines a browser bridge with a Chrome MV3 extension and exposes tools including browser_snapshot, browser_click, browser_type, browser_press, browser_scroll, navigation commands, browser_get_text, and browser_wait.

The extension turns a page into structured text containing its title, URL, main text, numbered interactive elements, and masked form fields. The agent addresses controls by number; screenshots are not sent through the model-facing pipeline. The bridge is registered in DSH’s local web profile, with the managed installation under ~/.dsh/dsh-browser and the extension under ~/.dsh/browser-extension. Local discovery uses /ext/bridge-config.

This is for browser tasks that depend on an existing login, session, or cookie state. It is not suitable for chrome:// pages, the Chrome Web Store, or workflows requiring screenshots or visual interpretation. The terminal surface includes shell-based installation and startup, while the browser surface reaches the selected Chrome tab: it can navigate, click controls, fill forms, and submit actions with the agent’s browser access. Password and payment-card values are masked before page data reaches the model.

dsh-browser documentation

How it behaves

dsh-browser controls the Chrome tab selected through its extension. browser_snapshot returns the page title, URL, main text, numbered controls, and masked form fields; delta: true returns only changes. browser_get_text reads a page region, including lazy-loaded or partial text. browser_click, browser_type, and browser_press operate controls and keyboard input. For browser_type, replace clears the field first. browser_scroll accepts viewport directions including up, down, top, and bottom. Navigation is provided by browser_navigate, browser_back, browser_forward, and browser_reload; browser_wait waits for page loading and rendering to settle.

Configuration and runtime

For local use, the extension discovers DSH through /ext/bridge-config. Loopback connections do not require a bridge token. DSH registers the bridge bundle in its local web profile and loads that profile at startup, so an already-running instance must be restarted after installation or an update. The default local address is http://127.0.0.1:3080. Remote deployment with --host 0.0.0.0 requires an address and bridge token.

Commands and paths

Start a managed checkout with pnpm start from ~/.dsh/dsh-browser, or start a cloned checkout with pnpm start from its repository root. The public release can be started with npx @deepseek-ai/dsh web. Both use port 3080 by default; --port <port> selects another port. The managed source is kept at ~/.dsh/dsh-browser; the built extension is copied to ~/.dsh/browser-extension.

Requirements and limits

The documented prerequisites are Node.js ^22.19 or >=24, Corepack/pnpm, and Google Chrome. The extension operates on normal http:// and https:// pages. Chrome-internal and protected pages, including chrome:// and the Chrome Web Store, cannot be read or operated. Password and payment-card values are rendered as •••• and do not enter the model-facing page data.

Written from the project's own documentation and kept in sync with it. Where the two disagree, the source is authoritative — read the README on GitHub

Same category

Alternatives to dsh-browser

38k

Four read-only tools over a WeKnora knowledge base: list knowledge bases, hybrid passage search, reassemble one document's chunks in order, and WeKnora's own cited RAG or ReAct-agent answer with a resumable session id.

No one-line install — this plugin lives inside a larger repository and publishes no npm package.

GitHub sourceGotoday

AI review

dsh-weknora

What it is — dsh-weknora provides four read-only tools to operate WeKnora knowledge base for DeepSeek Harness.

Who it is for — When using ReAct-agent in DeepSeek Harness to process WeKnora knowledge base, these tools can provide corresponding capabilities. If you do not use DeepSeek Harness for agent tool calls but rely on other integration methods, you do not need this plugin.

Watch out — Needs to provide API key or token. Will execute shell commands. No obvious pitfalls found.

The verdict — I will install it because it provides WeKnora knowledge base's RAG or ReAct-agent answer function.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Source

treg

superdesigndev

6.0k

Tool catalog for agents: search ~2,600 external endpoints (SEO and SERP, backlinks, social, people and company enrichment, ad libraries, scraping) by the task you want done, read each one's parameters and per-call price, then call it with the credential injected server-side. Ships the skill plus an MCP row that stays disabled until TREG_TOKEN is set.

Installed cleanly when we ran it

GitHub sourcePython24d ago

AI review

treg

What it is — Lets an agent find external endpoints by the job to be done, read params and per-call price, then invoke them with credentials injected server-side.

Who it is for — You are in DeepSeek Harness pulling backlinks, SERP, social, or company enrichment without Semrush-class seats. Skip it if secrets must stay on the caller host, or the environment cannot send traffic to an outside proxy.

Watch out — A clean-profile install completed and the harness registered the plugin. MCP stays disabled until TREG_TOKEN is set, and setup or runtime will run shell. Catalog usage is billed per call; newly verified accounts get a one-time $1.00.

The verdict — Skill is enough to shop the catalog. MCP stays off on my machine until the team is actually routing its own CLIs and skills through the same proxy.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Details

dsh-ops

T-Auto

5.3k

Provides bash, PowerShell 7, and high-performance Rust tools for DSH on Windows to reduce token usage.

DeepSeek Harness 0.2.2-rc.1 refuses the latest release

npm packageJavaScript11d ago

AI review

dsh-ops

What it is — An ops layer for DSH: extra bash, PowerShell 7, and Rust-backed tools, optional stock-Shell hiding, and on-demand loading so tool traffic stays small.

Who it is for — You are in Native or PTC on DeepSeek Harness, grinding bulk file edits or pdf and word reads, and oversized dumps plus error loops are chewing rounds. If the bill is almost all long reasoning with barely any tool calls, swapping this surface will not move cost.

Watch out — DeepSeek Harness 0.2.2-rc.1 rejects the newest release before install because the declared dsh range omits that version; 0.4.1 installed on the same host. The docs name 0.2.0-rc.2 and 0.2.1-rc.2 as supported hosts. It executes shell commands as part of the job.

The verdict — I would add it pinned to 0.4.1, or let dshmarketplace-cli choose—not latest.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Source

dsh-blender-plugin

sixtysevenlf

1.5k

Lets a DSH agent drive a local Blender over a direct TCP channel with 15 tools: viewport frames, custom-angle renders, Python execution, headless and job offload, transaction rollback, and a planning layer of 228 ops across 34 families.

Installs, but needs a build approved first

GitHub sourcePython18d ago

AI review

dsh-blender-plugin

What it is — Hooks a DSH agent into a live local Blender session over TCP, exposing 15 tools plus a 228-op planning layer.

Who it is for — You already run a DSH agent on a machine with Blender 4.x or 5.x—lofting, measuring a reference image, or auditing an assembly—and it still cannot grab a viewport frame or execute Python in that session. If your 3D path never attaches to a local Blender process, there is no session to drive.

Watch out — The sandbox got the package in, then pnpm blocked the build script and Harness never finished registration until that build was allowed by hand. The bundled Blender addon must be enabled first, with Node.js 20+ and Blender 4.x or 5.x. It executes shell commands.

The verdict — Worth it only on a box that already has Blender to connect; I would allow the blocked build first rather than chase a registration that never ran.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Details
1.4k

Anime Garden for DSH: search anime torrent resources from 動漫花園, 蜜柑计划, 萌番组 and ANi inside DSH — via chat, a sidebar panel with fansub and type filters, a seasonal calendar, and an in-app floating player.

No one-line install — this plugin lives inside a larger repository and publishes no npm package.

GitHub sourceTypeScript13d ago

AI review

AnimeGarden

What it is — Brings anime torrents from 動漫花園, 蜜柑计划, 萌番组, and ANi into DSH, searchable in chat or a sidebar, plus a season calendar and a floating player.

Who it is for — You are inside a DeepSeek Harness agent picking this season’s release and still leave the session to filter by fansub and type. If MCP or the HTTP API already feeds that index in another toolchain, skip this plugin.

Watch out — This listing has no one-line install; the plugin lives in a repo subdirectory and has to be wired in by hand. It runs shell commands. No sandbox run is on record, so there are no measured pitfalls.

The verdict — I’d add it only if torrent hunting already happens inside DSH most weeks.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Source

dsh-plugin

TencentCloudBase

1.2k

Tencent CloudBase backend for DeepSeek Harness — scaffold and deploy full-stack apps from chat, render query results as table cards with paging, sorting and CSV export, preview a deployment on its domain, and call the CloudBase MCP toolset (`mcp__cloudbase__*`) with device-code login.

Installed cleanly when we ran it

npm packageTypeScript2d ago

AI review

dsh-plugin

What it is — A DeepSeek Harness plugin that drives Tencent CloudBase from the thread: scaffold and deploy full-stack apps, present query results as table cards with paging, sorting and CSV export, preview a release on its domain, and call CloudBase MCP tools after device-code login.

Who it is for — You are shipping a full-stack app from DeepSeek Harness onto Tencent CloudBase and still lack an in-thread MCP bridge plus a DB / Storage / Auth panel. If CloudBase is not the runtime, this is the wrong door; if the Toolkit already lives in another AI IDE and DeepSeek Harness is only a second editor, the gap is not this plugin.

Watch out — On a fresh profile the sandbox install completed and DeepSeek Harness registered it. You still bring the CloudBase environment; login is device-code; shell commands run at runtime. Sensitive actions the agent proposes still need a human yes—the Toolkit does not supply judgment.

The verdict — A clean install is not a reason to add it; I would keep it only when query and release are actually supposed to happen in the DSH thread.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Details