Will the official team build my plugin in? DeepSeek Harness answers with an operating system analogy

Tianyi Cui responded on X to a question plugin authors keep asking: which features will the official team build in? If the official team builds it, is there still any point in third-party plugins?

He used smartphone operating systems as an example. Systems come with built-in alarm clocks and calculators to meet basic needs. Phone manufacturers would rather see better alarm clocks and calculators emerge in the ecosystem, and they want third parties to use system APIs to build apps the people who invented phones never imagined. DeepSeek Harness has a similar idea: provide stable foundational capabilities and build in the general features most people need. Under the "everything is a plugin" design, built-in features can also be replaced by third-party plugins. The official team hopes more community works become users' everyday choices, with the goal of making DSH the operating system of the Agent era.

The official article shared some scale numbers. DSH v0.1 launched on August 13, and the desktop version was released on September 29. As of November 4, DSH on the official API reached 1.54 million daily active users, consuming about 92 trillion tokens per day. There are about 35,000 third-party plugins, and about 60% of active users have created or installed at least one plugin. Plugins are no longer a niche toy.

The official article listed several plugins. To install them, enter the package name on the DSH plugin installation page.

Creator workbench, package name dsh-media-desk, by woniuniuniu. It puts capturing ideas, writing scripts, making covers, and storyboarding into one workbench. The author's description: wake up, open DSH, the workbench starts in seconds, and then get to work.

Design plugin, package name dsh-guizang-design, by op7418. It combines common design and copy-polishing skills into a plugin that can turn articles into web pages, image cards, or covers. Finished works are saved in My Work, and clicking Make a Similar One lets you keep editing. The author says the plugin system offers more room than skills, and works, content series, and writing styles can be preserved and continuously iterated.

Context insight, package name dsh-context, by bowenliang123. It shows how context is composed, how it changes, and the details of multi-agent collaboration. The author says the open and flexible plugin mechanism keeps the agent's context and the harness itself open source, and hands control back to users.

Paper reading, package name agentero-dsh, by Phil-Fan. It puts paper reading, highlighting, translation, and notes in one place, and lets the agent help organize them. The author says the plugin mechanism connects reference management, translation, and notes into one literature tool.

Beyond these examples, heavier plugins have already appeared in the community. The professional design tool OpenPencil joined DSH during its beta, allowing real .op documents to be previewed, inspected, and edited in the conversation, with an interactive canvas and hosted editor, and the agent can directly operate the design canvas through tools. Project address: https://github.com/ZSeven-W/dsh-openpencil

The game workshop dsh-game-studio turns "make a mini-game with one sentence" into a plugin. It includes 2D top-down and 3D voxel templates, games preview in real time in the right sidebar, and the agent playtests after writing, checking frame rate, catching errors, and looking at screenshots. AI art candidates can be swapped by number, with version rollback, LAN mobile playtesting, and export to a zero-dependency single HTML file.

Some people are pushing DSH toward a personal AI OS. Captain｜Yufei built DeepSeek Dock, making the sidebar look like the macOS Dock and gathering common entry points. In the video you can see him repeatedly adjusting the sidebar interaction.

Some users also pointed out the interface problem once there are more plugins: plugins and sessions are crowded into the sidebar, and information becomes cluttered. Some hope plugin entries that need an interface can be moved to the top.

Someone also built a plugin promo video tool called dsh-reel, which can generate horizontal videos, video covers, and vertical promo images from plugin introductions, screenshots, and examples.

The official team's next step will be improving the stability and compatibility of the plugin API, strengthening plugin security, and building an official plugin marketplace so developers can publish plugins and users can find them. The official team is also preparing a plugin author support program with token grants, technical discussions, and promotion of works. Specific details will be announced once community needs are clearer. Plugin authors who want to participate can fill out this form: https://trtgsjkv6r.feishu.cn/share/base/form/shrcnkRrxNyri4bOF9n48BIYWbb

The concerns in the comments are more specific.

Someone asked how to solve mixed quality and difficulty choosing. As the number of plugins grows, search, categorization, reviews, and security review will all become basic capabilities of the marketplace. An official marketplace is one answer, but the marketplace itself will also face ranking and trust issues.

A user said many plugins are actually filling in features the official product lacks, not doing better than features the official product already has. This time window really exists. Once the official version builds something in, plugins either have to become stronger or migrate to narrower scenarios.

A developer worried about the official team crushing competition: if the official team builds the same feature, it could be overwhelming for third parties. They hope the official team draws clear boundaries so developers know what the official team will build and what is left to the community. Boundaries will change with versions, and stable APIs and migration mechanisms are more practical than one-time promises.

A user said that rather than worrying about whether the official team will build the same thing, they care more about whether the plugin API can remain stable long term. Once users get used to something, stability matters more than novelty.

Security is also a hard problem. Not everyone can judge whether an open-source plugin is malicious. If an official marketplace is built, permissions, dependencies, update sources, and review records all need to be made clear.

Plugin conflicts and reinventing the wheel were also mentioned. Someone suggested using complex network clustering to test duplicate features and conflicting modules. The idea is very engineering-minded, but hard to implement. At the very least, the official team could first provide dependency warnings, permission warnings, and conflict detection.

The most concrete question is data migration: if you switch to another paper reading plugin, are your original annotations still there? If data cannot be taken with you, replaceability is only nominal. An operating-system-level ecosystem must answer this question.

Whether comparing DSH to the operating system of the Agent era holds up does not depend on the number of plugins. The key lies in several mechanisms: API stability, permissions and security, marketplace discovery, data portability, and replacement without lock-in. The real room for third parties is in vertical workflows, professional tool integration, and personal preferences. Filling in features the official product lacks has a time window; once the official team builds them in, plugins either have to become stronger or narrower. Official boundary-drawing helps, but boundaries will change. Stable APIs and migration mechanisms are more practical.