AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Prime Big Deal Days · Oct 6–7Offer from Amazon

Get tech for your team delivered free — and shop member deals

  • Fast, free delivery on millions of items
  • Access to Prime Big Deal Days deals on October 6–7
  • Prime Video, Amazon Music and more included
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

DeepSeek has opened public preview of DeepSeek Harness, an open-source agent workspace offered as a desktop app for macOS and Windows and as a web UI. The company describes plugins for everyday work, coding, research and scheduled tasks; availability details, system requirements and preview limitations are not specified in the supplied announcement.

DeepSeek says DeepSeek Harness, an open-source workspace for AI agents, is now in worldwide public preview, with a desktop app for macOS and Windows and a web interface that can be launched from code. The release gives users a way to apply the company’s agent tools to tasks such as document work, coding and research, while allowing plugins to add or change capabilities.

DeepSeek describes Harness as built on Cordis’s “everything is a plugin” architecture. Its announcement says plugins can extend the tools, skills and interface, and users can install plugins or create them through chat in a feature called Creator mode. The source presents the product as open source, though it does not set out licensing terms on the supplied page text.

The listed uses include organizing files, analyzing data, drafting documents and creating slides, alongside exploring code repositories, fixing bugs, building features and running tests. For research, DeepSeek says the system can find information, verify facts and provide citations. The announcement also describes previewing work and refining it through conversation.

For ongoing work, DeepSeek says users can enable a Scheduled tasks plugin to start tasks at a chosen time or run recurring work, and inspect progress. It also advertises developer tools that expose execution traces and runtime information for troubleshooting tool calls and task execution. These are product descriptions from DeepSeek; the supplied source does not provide independent performance testing or user results.

At a glance
announcementWhen: Announced as available now in worldwide…
The developmentDeepSeek has made its open-source Harness agent workspace available in worldwide public preview, including desktop versions for macOS and Windows.

A Desktop Workspace for Agent Tasks

The release puts DeepSeek’s agent interface into a desktop workflow as well as a web UI, positioning Harness for work that may involve local files, code projects or repeated tasks. That could make it relevant to people who want an agent to operate across several kinds of work rather than use a separate interface for each task. The announcement does not establish how the app accesses files or what permissions it requires.

Its plugin-based design is also central to the product’s pitch: users and developers can adapt available tools and interface elements, rather than relying only on a fixed set of built-in functions. Public preview means the software is being offered for broader use, but it does not by itself confirm stability, feature completeness or suitability for sensitive work. Those questions matter for users evaluating an agent that may handle documents, data or code.

Amazon

macOS AI agent desktop app

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

How DeepSeek Says Harness Works

The announcement presents Harness as a single environment for everyday tasks, coding, research and background work. Users can describe what they want to do, use slash commands, refer to files or sessions, and inspect results. DeepSeek’s page also names DeepSeek-V4.1-Flash in the interface description, but the supplied material does not explain the model’s role, availability or relationship to the desktop release.

DeepSeek gives two routes for using the web interface: install Node.js and run it with the command npx @deepseek-ai/dsh web, or clone the project repository and follow its setup instructions. These instructions concern the web UI and source installation; the announcement separately says desktop apps are available for macOS and Windows. It does not provide download links or version numbers in the supplied text.

“DeepSeek Harness is now in public preview worldwide and open source, with composable plugins that extend what agents can do.”

— DeepSeek, in its Harness announcement

Amazon

Windows open-source automation tool

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Preview Details Still To Come

The supplied announcement does not give a release date, version number, system requirements or direct desktop download links. It also does not explain whether every feature is available on both operating systems, which plugins are included by default, or whether users need a DeepSeek account. DeepSeek calls the software open source, but the material provided here does not specify the license or how the desktop packages correspond to the repository.

Other operational details remain unaddressed, including where task processing takes place, what data may be sent to online services, how local-file access is controlled, and whether scheduled tasks continue when the app is closed. The source also makes no independent claims about accuracy, reliability or security. Those details should be checked in product documentation before users rely on Harness for sensitive files or consequential work.

Amazon

AI research and coding workspace software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Where Users Can Check Availability

DeepSeek’s announcement says Harness is ready to use in public preview and points developers to its project repository and web-launch instructions. Users interested in the desktop apps will need to check DeepSeek’s official Harness page for current macOS and Windows download information, since the supplied source text does not include package links or installation steps.

Further details to watch for include platform-specific requirements, plugin documentation, licensing information and guidance on data handling. DeepSeek has not provided a timetable for changes to the preview or a full release in the supplied announcement, so no later release milestone can be confirmed.

Amazon

file organization and data analysis tools for Mac

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

What is DeepSeek Harness?

DeepSeek Harness is an open-source agent workspace that DeepSeek says can be used as a desktop app or through a web UI. The company describes it as a tool for tasks including coding, research and document work.

Which operating systems are supported?

The supplied announcement identifies desktop apps for macOS and Windows. It does not give specific supported versions or hardware requirements.

Is DeepSeek Harness publicly available?

DeepSeek says the product is available now in worldwide public preview. The source material does not include desktop download links or say whether registration is required.

How can developers launch the web UI?

DeepSeek’s page says to install Node.js and run npx @deepseek-ai/dsh web, or clone the project repository and follow its setup instructions.

Can users add their own capabilities?

DeepSeek says users can install plugins or create them through chat in Creator mode. The announcement does not list the complete set of plugins or explain their permissions.

Source: hn

HALLOWEEN

Halloween Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

Understanding Cumulative Attention Loads In School Software For K-12

A new approach measures the total attention burden of school software portfolios, helping districts make informed procurement decisions amid rising screen-time concerns.

Show HN: Microsoft Releases Flint, A Visualization Language For AI Agents

Microsoft announces Flint, a new visualization language designed for AI agents, aiming to improve reliability in generating data visualizations.

Show HN: Open-source Engine Running Gemma 4 26B In 2 GB RAM On Any M-series Mac

A new open-source engine, TurboFieldfare, enables running Gemma 4 26B AI model on any M-series Mac with only 2 GB RAM, using Swift and Metal.

Next Mythos-Class Model Released By September 1, 2026?

Speculation suggests a new Mythos-Class model may be released by September 1, 2026, amid rising interest and market signals, but official confirmation is pending.