Aurora Mobile Limited (NASDAQ: JG) (“Aurora Mobile” or the “Company”), a leading provider of customer engagement and marketing technology services, announced that Modellix.ai, its One Unified Platform for All AI Media Generation, has released dsh-modellix in beta — a plugin for DeepSeek Harness, the open-source AI coding agent that has drawn nearly 200,000 GitHub stars within two weeks of its debut (as of August 27, 2026). The plugin lets Harness users call LLMs through Modellix’s gateway — including models billed at $0 — using a single Modellix API key.

DeepSeek Harness, open-sourced on August 13, 2026, is built on a plugin-first architecture — its motto is “Everything is a Plugin” — that lets developers extend their coding workflow with third-party providers. dsh-modellix integrates Modellix’s LLM gateway into Harness’s native model selector, with three independently toggled capabilities: LLM, Design, and Web tools.

Marketing Technology News: MarTech Interview with Mark Listes, CEO @ Pendulum Intelligence

A $0-per-call way in
The headline capability is the entry point at zero per-call cost: Harness users can switch their conversation model to “modellix-ai/free-llm” — a model that automatically routes requests to current state-of-the-art open-source LLMs — or to “zai/glm-4.7-flash”, also billed at $0. No new sign-up flow beyond a Modellix account and API key, and these models are billed at $0. Developers who want more can switch to any of the 28+ models on the gateway without changing keys — see the full catalog with pricing at https://www.modellix.ai/llm.

“AI coding agents are becoming the default front door to AI development, and the ecosystem around DeepSeek Harness is growing remarkably fast,” said Mr. Weidong Luo, Chairman and Chief Executive Officer of Aurora Mobile. “dsh-modellix is an early beta, and our goal is simple: a developer using Harness should be able to try the latest models — including those billed at $0 — in under two minutes, with one key that also unlocks image, video, and audio generation on the same platform. We’re shipping it early to learn what the community actually wants, and iterating from there.”

Marketing Technology News: How MarTech Is Enabling Autonomous Brand Engagement Across Channels?

What the plugin provides

• LLM — syncs Modellix’s live model catalog into Harness’s model selector, covering 28+ models across OpenAI, Anthropic, Google, xAI, GLM, Qwen, DeepSeek, and Kimi. Supports OpenAI-compatible Chat Completions and Responses, and Anthropic-compatible Messages.
• Web — routes Harness’s native web_search and web_fetch tools through Modellix’s tool endpoint.
• Design — enables image, video, and audio generation inside Harness’s Design view (available as a separate toggle).

The LLM capability is functional in this beta; Design remains in preview as Modellix continues to harden the experience across upstream model schemas. As a beta, the plugin’s APIs and behavior may change before the full release.

One key, three protocols, no lock-in
Modellix’s LLM gateway at llm.modellix.ai is protocol-compatible with the OpenAI and Anthropic ecosystems: existing SDK and client configurations — Codex, Claude Code, Cursor, and others — can point at Modellix with a base-URL override, with no code rewrite. Latest Model IDs (e.g., “~openai/gpt-latest”) keep developers on the current flagship of each series automatically. The gateway is a one-stop path to models from multiple vendors without managing separate accounts.