DeepSeek Harness Creator Mode lets users generate AI plugins from text prompts

DeepSeek Harness Introduces Creator Mode for Plugin Generation

DeepSeek has unveiled Creator Mode for its DeepSeek Harness platform, a feature that enables users to generate plugins from text prompts using natural language prompts.1

1

2

2

The Hangzhou-based company announced the launch on Monday, marking a significant step in AI agent development by making plugin creation accessible through conversational interfaces.2

2

The new Creator Mode allows users to build custom plugins by simply typing instructions in everyday English, eliminating the need for extensive coding knowledge.2

2

4

4

DeepSeek demonstrated the capability with examples including a digital sticky note wall, a poster design tool, plugins for 3D model rendering, a video editing workbench, and a Minecraft-like video game.1

1

3

3

Understanding the Open-Source AI Harness Platform

DeepSeek Harness operates as an open-source framework built around the philosophy that "everything is a plugin."1

1

3

3

While AI models excel at generating responses and determining courses of action, they require a harness to execute those actions. Harnesses connect LLMs to tools and systems capable of performing real-world tasks such as scheduling calendar appointments or making phone calls.1

1

3

3

When combined, an AI model and harness create AI agents that can autonomously handle complex operations.

The platform describes its infrastructure as open-source, reusable, and composable, with evolving core plugins and APIs currently in preview.2

2

Users can launch the Web UI by installing Node.js and running commands with npx, or clone the full source from the project's repository.2

2

This open-source approach positions DeepSeek Harness as a flexible foundation for developers building sophisticated AI applications.

Text-to-Plugin Generation Capabilities and Implementation

The text-to-plugin generation feature enables users to create custom AI plugins through chat within DeepSeek Harness, extending tools, skills, and interfaces without altering the original program.2

2

3

3

DeepSeek published demonstration prompts in both English and Chinese, allowing users to experiment with the system.1

1

3

3

However, the company cautioned that some plugins shown in demonstrations underwent several rounds of conversation-based refinement, meaning single prompts may not perfectly reproduce the showcased results.1

1

3

3

DeepSeek's AI agent targets developers rather than casual consumers, focusing on professional use cases.2

2

4

4

The company formed the Harness team earlier this year specifically to develop AI agents that rival AI coding services, competing with offerings from US firms including Anthropic PBC.2

2

4

4

Strategic Positioning in the AI Agent Race

DeepSeek's Creator Mode launch comes as leading AI developers in the US and China race to build agents capable of handling tasks such as coding and financial research for professionals and ordinary consumers.2

2

4

4

The company rolled out an upgraded coding assistant on Saturday, designed to step in on problems early and demand minimal human oversight.2

2

4

4

In its latest funding round, DeepSeek raised approximately 100 billion yuan ($15 billion), ranking among the biggest raises in China as the company prepares for an initial public offering.2

2

Harness also played a role in DeepSeek's October model update, where the company optimized an updated V4.1-Flash model specifically for DeepSeek Harness and recommended upgrading to Harness 0.2.1-rc.1 or later.2

2

This substantial investment signals confidence in DeepSeek's approach to AI agent development and its potential to compete with established Western AI companies in the rapidly evolving market for autonomous AI systems.