DeepSeek open-sources agent runtime Harness with plugin-based architecture
DeepSeek Harness, an open-source agent runtime, was quietly released on August 13, 2026, and quickly gained traction, reaching 186,000 stars and 20,000 forks within ten days. The project’s core innovation lies in its plugin-based design, where every component—from the model adapter to the UI—can be swapped independently, built on Cordis, a framework previously used in the Koishi chatbot project…
Key points
- DeepSeek Harness open-sources an agent runtime with **152+ independently swappable plugins**, including UI panels and the agent loop
- Supports **40+ model providers** and **OS-level sandboxing** (Linux/macOS/Windows) with **append-only session logs**
- Project is a **developer preview**—not for end users—with **compatibility-breaking changes** expected
The runtime supports 40+ model providers, allows delegation to competitors’ agents, and enforces OS-level sandboxing (using bwrap, Seatbelt, or Landlock) by default. Session logs are append-only, ensuring transparency. While labeled a developer preview, the project warns of compatibility-breaking changes and lacks polished end-user features. The npm package (@deepseek-ai/dsh) is already installable, with 152+ plugins documented in its default configuration. Early adopters—particularly those building agent infrastructure—are drawn to its modularity, though it remains not production-ready for general use.
What I’ve Learned About DeepSeek Harness
KDnuggets · 24 September 2026
Loading the full article…
This text was published by KDnuggets and written by Shittu Olumide. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- ChatGPT, Gemini and Claude offer personalization settings · 1 src
- Book teaches engineers to integrate LLMs via API for business workflows · 1 src
- Zapier blog lists six best AI resume builders · 1 src
- Google says Gemini autonomously accessed three companies, then stopped itself · 2 src
- Meta’s Muse leaks internal files via prompt injection, developers say · 1 src
Comments
via GitHub Discussions