The company launched Harness on August 13 and introduced a desktop version on September 29. The application functions as an AI agent runtime, allowing users to carry out activities such as programming, document analysis, and research with limited manual intervention.
DeepSeek also reported significant usage across its infrastructure. According to the company, Harness users process approximately 92 trillion tokens daily through its official API. That works out to nearly 60 million tokens per daily active user, although the company did not provide a breakdown of usage across different activities or users.
The figures reflect activity measured through DeepSeek's official API rather than an independently verified assessment of the application's entire user base. They also do not establish how many people use Harness through other connections or configurations.
Alongside the increase in users, Harness has attracted a growing community of third-party developers. DeepSeek says approximately 35,000 plugins have been created for the application, allowing users to add functionality beyond its standard features.
About 60% of active users have either developed or installed at least one plugin, according to the Harness team. That level of participation suggests extensions are becoming an important part of how users interact with the application, although DeepSeek has not disclosed how frequently individual plugins are used or how their functionality varies.
To support that activity, DeepSeek is preparing an official marketplace where developers will be able to distribute Harness plugins. The company also intends to introduce a token-based incentive initiative for developers and a separate support program for plugin creators.
The proposed support measures could include AI usage credits, technical assistance, and promotional opportunities, according to additional reporting. However, DeepSeek has not provided a launch schedule or detailed eligibility requirements for these initiatives.
The expansion comes as AI companies develop agent-based applications that can perform multistep tasks rather than respond only to individual prompts. Harness is positioned as an open-source option in a market that also includes agent products from OpenAI and Anthropic.
DeepSeek's reported token consumption illustrates the computing requirements associated with these applications. Agent workflows can involve repeated interactions with AI models as software processes information, executes tasks, and evaluates results. The company has not disclosed the infrastructure costs associated with Harness usage.
Several commercial details also remain unclear. DeepSeek has not specified the application's pricing structure, licensing terms, potential revenue-sharing arrangements for its planned marketplace, or which models Harness uses by default.
For now, the company's disclosures point to rapid reported adoption and substantial participation in its plugin ecosystem. The next development will be whether DeepSeek can translate that activity into a more established developer platform as it introduces official distribution and support tools.
This analysis is based on reporting from Seeking Alpha.
Image courtesy of DeepSeek.
This article was generated with AI assistance and reviewed for accuracy and quality.
