DeepSeek’s Agent Tool Hits 1.54 Million Daily Users

Human Verified 0

Share to: , and

A small robot assembling glowing networked plugin tiles in a dark teal digital workspace

DeepSeek’s open-source agent platform has grown far faster than its own team expected. The Harness team said on the evening of 5 November that daily active users reached 1.54 million on 4 November, up from 271,000 on 30 September, with those users consuming about 92 trillion tokens a day. Around 35,000 third-party plugins have appeared since launch, and about 60% of active users have created or installed at least one. The team also said it is preparing an official plugin marketplace and a support programme for plugin authors, with no launch date given.

## Five weeks, five times the users

The trajectory is the story. Daily users averaged 271,000 over the seven days to 30 September, then reached 429,000 by 7 October and 1.54 million on 4 November — about 5.7 times the September figure, or more than a million new users in just over a month.

The count covers only people using Harness through DeepSeek’s official API, so anyone running it through another model connection sits outside the number. One caution applies to the comparison: September’s figure is a seven-day average, while November’s is a single day, so the growth is real but the two numbers are not measured the same way.

## Plugins are the product now

Harness is built around the idea that everything is a plugin — tools, memory, panels and model connections are all replaceable parts rather than fixed code. That design is paying off in volume: about 35,000 plugins from outside developers have appeared since the August launch.

Roughly 60% of active users have built or installed at least one plugin. [DeepSeek showed off the idea on 2 November](https://geroundtable.com/deepseek-creator-mode-builds-own-plugins/), when it published a video of the agent writing, installing and testing its own plugins from plain-language requests. An official marketplace would give that sprawling catalogue a central home, replacing the community indexes developers rely on today.

## A marketplace, and help for authors

The team said it is preparing an official plugin marketplace alongside a support programme for authors, offering token incentives, technical exchanges and featured placement, with the details still shaped by community feedback. No launch date was given for either the marketplace or the author programme.

It also promised to stabilise the plugin interface and strengthen security, two commitments that matter because plugins are code that can run commands and touch files. [The direction was set out on 31 October](https://geroundtable.com/deepseek-super-agent-autonomous-coding/), when the team published its super-agent design article describing agents that keep working across sessions and juggle several tasks at once.

## The catch: it is still a preview

Growth has outrun the safety work. Harness remains a developer preview, and DeepSeek warns that the software has not undergone a security audit, that plugins can damage files or leak credentials if a prompt goes wrong, and that breaking changes are coming.

The October bridge for extensions written for Claude Code ran some of them without a sandbox, which shows why the promised security work matters. [Independent testing has already shown DeepSeek’s models can behave unpredictably under pressure](https://geroundtable.com/deepseek-fangcun-safety-tests/), so a marketplace that makes installing other people’s code easier will need the guardrails to arrive first.

A laptop on a desk at dusk showing colourful interlocking plugin modules on its screen

## Our Read

The numbers change what Harness is. A month ago it looked like a promising open-source experiment; at 1.54 million daily users it is a platform with an ecosystem, and the 60% plugin adoption rate says users are not just trying it — they are extending it.

That is the moat DeepSeek is building: not a better model, but a workspace users reshape themselves. [The timing is no accident either: Moonshot launched hosted Kimi agents for enterprises on 5 November](https://geroundtable.com/kimi-hosted-agents-enterprise-launch/), the same day DeepSeek disclosed its figures, so the open agent race now has two Chinese contenders moving in the same week.

The comparison that matters is with the closed rivals in the agent race. [Perplexity’s always-on Teammates charge $200 a month and run inside its own cloud](https://geroundtable.com/perplexity-teammates-ai-coworkers/), while DeepSeek gives the harness away free and charges only for model use through its API.

Openness is the cheaper bet, but it carries the cost DeepSeek itself flags: unaudited code, breaking changes and a plugin supply chain nobody has yet secured. If the marketplace lands with real review and sandboxing, the bet pays; if it lands as a bare directory, the first serious plugin incident will define the project instead.

## Why It Matters

Agent platforms live or die on their ecosystems, and Harness has just shown its ecosystem is alive: 35,000 plugins, most users building or installing them, and a marketplace on the way. For developers, the practical question is whether plugins built today survive the breaking changes DeepSeek has warned about.

The other question is whether the marketplace turns one-off demos into tools other people can trust. For the wider industry, the figures are the first hard evidence that an open-source agent harness can reach mass usage in under three months — and that the agent race is no longer only about who has the smartest model.

## Background

[DeepSeek released Harness in August under the MIT licence, shipped desktop apps with the version 0.2 preview in late September, and closed a $16 billion funding round led by state capital on 1 November.](https://geroundtable.com/deepseek-16-billion-second-round/) Its public repository has passed 270,000 stars.

Source: [Tech in Asia report](https://www.techinasia.com/news/deepseek-harness-hits-154-million-daily-users-468) and [Asia Daily report](https://asiadaily.org/news/17427/)

Share to: , and

Faheem R. Khattak
Founder & Senior Editor

Senior automation engineer and AI marketing journo covering LLMs, algorithms, analytics, core updates, Bing search, and technical SEO.

✓ 10+ Years Covering Google Search & AI

0 Comments

Leave a Reply

Optional Take Tag:

🛡️ Spam protection: Honeypot + Time-Trap Active

This site uses Akismet to reduce spam. Learn how your comment data is processed.