On August 13, DeepSeek released a developer preview of DeepSeek Harness, its agent harness, with source code included and an MIT license. The site opens by stating its thesis: “Agent = Model + Harness.” The model is the soul of an agent — but the layer that lets an agent understand its environment, use tools, and keep working in real-world settings is the harness. The launch pulled 747 points on Hacker News, making it one of the most-watched developer tool releases of the week. Installation is one command — npx @deepseek-ai/dsh web (it requires Node.js) — and the source lives at the deepseek-ai/deepseek-harness repository on GitHub.
Everything Is a Plugin: the Cordis Kernel
The harness is built on the Cordis kernel, which handles plugin mounting, unmounting, and dependency management. Models, tools, skills, sessions, sandboxes, storage, loops, scheduling, and the UI are all plugins. Developers can select, swap, or extend any capability in configuration without touching the DeepSeek Harness source code. In practice that means the differences between “our agent” and “the stock agent” can live entirely in a config file and a plugin directory: swap the model provider, replace the sandbox, add a scheduling loop, and your fork stays mergeable upstream. The design makes it feel less like a single CLI product and more like an assemblable operating system for agents. Cordis itself is a mature kernel with an existing community, which shortens the path from idea to working plugin.
Replayable Run Logs
The second pillar is traceability. Everything the model sees — system prompts, reasoning, tool calls and results, subagent scheduling, context injections — goes into an append-only session log. Resume, fork, search, and replay all operate on the same event stream. For debugging and evaluation, that is a far more complete experience than reading a transcript after the fact: a failed run can be replayed from any point with a different tool result or prompt, and a change in behavior can be bisected across the event history. It also turns “why did the model do that” from an argument into a question you can replay step by step. It is the same idea version control applies to source code, applied to agent behavior: the run is a log of events, so any question about the run becomes a query over that log.
Four Runtime Modes
The preview ships four modes. Standard is the full coding agent: file editing, shell, search, skills, planning, goals, subagents, and workflows. Code exposes tools through a Code Mode SDK so the model can chain multi-step operations into one TypeScript program that runs in a single pass — fewer round trips, fewer chances for the model to lose the plot between calls. Minimal strips it down to persistent bash and str_replace_editor, aimed at benchmarking models in a bare environment; it is effectively a reproducible harness for model evals. Creator adds runtime inspection, in-memory plugin testing, and preset authoring. Four modes, four use cases: daily development, batch automation, benchmarks, and plugin development. All four share the same plugin system, so a skill learned in Minimal carries into Standard unchanged.
The Harness Race Is On
The industry signal here is clear: model companies now treat the harness as a product line of its own. Third-party toolchains used to fill this layer; DeepSeek has stepped in directly, and with an MIT license and a plugin ecosystem from day one — community plugins are already gathering under the dsh-plugin tag, and the Cordis kernel has a corresponding paper. The company is honest about the caveats: the preview is still being tested, and core plugins and APIs will continue to evolve, so production pins should expect breakage. Even so, the direction matters more than the version number. If the model is the soul of the agent, the harness is the body — and until now, most teams building on frontier models had to accept a closed body they could not inspect. An MIT-licensed, fully swappable layer changes the default: the cost of getting in early has never been lower, and the cost of being locked in just became a choice rather than a given.
Sources
- DeepSeek Harness — official developer preview
- deepseek-ai/deepseek-harness — GitHub
- DeepSeek Harness developer preview — Hacker News
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.