# DeepSeek's open-source agent app Harness reaches 1.54 million daily active users > DeepSeek's open-source agent app hit 1.54 million daily active users on November 4, less than three months after launch. Daily token consumption reached roughly 92 trillion, a volume the company absorbs directly on its own infrastructure. - Type: AI news - Page: https://completeaitraining.com/news/deepseeks-open-source-agent-app-harness-reaches-154-million/ - Published: 2026-11-09 - Updated: 2026-11-09 - Author: Jeroen Erne - Jobs: Creatives, IT and development, Marketing, Product development, Sales, Science and research - Reading time: 2 min DeepSeek's open-source agent app Harness reached 1.54 million daily active users on November 4, less than three months after its August 13 debut, according to figures the company shared with Bloomberg. The growth curve is steep: a 7-day average of 271,000 users on September 30 jumped to 429,000 by October 7, then surged past 1.5 million in under a month. Daily token consumption through DeepSeek's hosted API hit roughly 92 trillion, a volume that runs directly against the company's own infrastructure. The numbers are self-reported and unaudited. No third party verifies the user count or the token figure. But the disclosure lands during a visible run of Chinese-origin open-source releases and arrives while Anthropic and OpenAI keep Claude Code and ChatGPT's agent surface closed. Bloomberg framed the move as DeepSeek continuing to "challenge Anthropic PBC and OpenAI with an entire swath of applications built around open-source AI technologies." #### The plugin ecosystem is already forming About 35,000 third-party plugins now ride on the agent platform. DeepSeek said roughly 60% of active users have created or installed at least one plugin. The company is preparing an official plugin marketplace and an author-support program, and has already started distributing tokens to plugin developers. That effectively funds the ecosystem before the storefront opens, though no launch date was given. For developers building on agent frameworks, this signals a real distribution channel. A marketplace with developer payouts creates incentives that closed-source competitors have not yet matched at this scale. For professionals evaluating where to invest time learning agent workflows, the [Deepseek Courses](https://completeaitraining.com/tag/deepseek/) page offers structured training on the platform's capabilities. #### What the token volume actually means Ninety-two trillion tokens per day is a number that demands context. It represents the computational load of serving 1.54 million users running agent tasks - not simple chat completions, but multi-step workflows that chain API calls. DeepSeek is counting this traffic against its own hosted API, meaning the company absorbs the inference cost directly. The token figure also functions as a proxy for usage intensity. High token consumption per user suggests the agent is doing real work rather than casual experimentation. For IT and development teams tracking total cost of ownership for AI tooling, these numbers provide a rough benchmark for what open-source agent deployment looks like at scale. #### Open-source agents versus closed ecosystems The contrast with Western competitors is structural. Anthropic's Claude Code and OpenAI's ChatGPT agent features remain closed, with usage gated behind the companies' own interfaces and pricing. DeepSeek's approach - an open-source agent with a hosted API and a third-party plugin layer - mirrors the playbook that built ecosystems around operating systems and browsers. That architecture matters for product development and sales teams. An open agent that supports plugins can be customized for industry-specific workflows without waiting for a vendor roadmap. For those exploring how to integrate such tools into existing processes, [AI Agent Courses](https://completeaitraining.com/tag/automation/) cover practical implementation patterns across business functions. #### Why this matters for creatives, developers, and product teams For creatives and marketers, a plugin marketplace attached to a widely-used agent creates new surfaces for content tools, brand integrations, and workflow automation that do not depend on a single vendor's feature schedule. Developers gain a distribution channel with built-in monetization before the storefront even launches. Product and sales leaders should track whether an open-source agent ecosystem can deliver customization speed that closed competitors cannot match - and at what infrastructure cost. Science and research teams get a real-world dataset: 1.54 million daily users generating 92 trillion tokens through an open-source stack is a signal worth watching, even if the numbers come from the company itself. --- This is the Markdown form of https://completeaitraining.com/news/deepseeks-open-source-agent-app-harness-reaches-154-million/. The whole site is mapped at https://completeaitraining.com/llms.txt.