Everything Is a Plugin: Inside DeepSeek Harness, the Week the Scaffolding Became the Product
Last Updated on August 25, 2026 by Editorial Team
Author(s): Shrashti Singhal
Originally published on Towards AI.
Everything Is a Plugin: Inside DeepSeek Harness, the Week the Scaffolding Became the Product
A year ago, “the harness is the product” was an argument. On August 13, DeepSeek turned it into a strategy: an MIT-licensed agent harness where the models, the tools, the sandbox, and the loop itself are all swappable, and 90,000 GitHub stars arrived in two days. Here’s what it actually is, how it differs from the harnesses we’ve been building, and what the fine print says.
The article explains DeepSeek’s “dsh” (DeepSeek Harness) as a developer-preview agent runtime where the core loop, model adapters, sandbox, session/logging, UI, and even sub-agent providers are all treated as MIT-licensed plugins configured via YAML—effectively shifting “harness is the product” from slogan to implementation. It details the machinery (Cordis composition, event-driven lifecycle, and an append-only session log as the run’s source of truth), the tool pipeline and permission model (approval vs sandbox reachability), and how traditional harness architectures differ when the loop and other “fixed” components become swappable interfaces. Finally, it weighs early fine print from the first two weeks—token overhead, limited/fragile third-party integrations, benchmark methodology opacity—and concludes with practical guidance on when to adopt dsh versus wait, plus key takeaways (log-is-the-run, separated approval/sandbox questions, tool batching/code mode, and per-session effort-tunable routing), arguing that regardless of outcomes, the harness layer has become a battleground.
Read the full blog for free on Medium.
Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.
Published via Towards AI
Towards AI Academy
We Build Enterprise-Grade AI. We'll Teach You to Master It Too.
15 engineers. 100,000+ students. Towards AI Academy teaches what actually survives production.
Start free — no commitment:
→ 6-Day Agentic AI Engineering Email Guide — one practical lesson per day
→ Agents Architecture Cheatsheet — 3 years of architecture decisions in 6 pages
Our courses:
→ AI Engineering Certification — 90+ lessons from project selection to deployed product. The most comprehensive practical LLM course out there.
→ Agent Engineering Course — Hands on with production agent architectures, memory, routing, and eval frameworks — built from real enterprise engagements.
→ AI for Work — Understand, evaluate, and apply AI for complex work tasks.
Note: Article content contains the views of the contributing authors and not Towards AI.
Related posts
Recent Posts
