Featured starting point
DeepSeek Harness Explained: Plugins, Profiles, and the Agent Runtime
A source-backed introduction to DeepSeek Harness (dsh), its Cordis plugin architecture, profiles, bundles, and developer-preview boundaries.
Browse all 9 articles in this editorial series. A source-backed guide to DeepSeek Harness, from first run to Cordis architecture analysis.

Featured starting point
A source-backed introduction to DeepSeek Harness (dsh), its Cordis plugin architecture, profiles, bundles, and developer-preview boundaries.
Suggested learning path
01 → 09
A source-backed introduction to DeepSeek Harness (dsh), its Cordis plugin architecture, profiles, bundles, and developer-preview boundaries.
Follow the verified npm and source-checkout paths for DeepSeek Harness, then configure a model and workspace in the Web UI.
Trace DeepSeek Harness from profile composition to a model step, using the repository architecture guide and package maps as a reading plan.
Use the verified Web UI workflow to configure a model, select a workspace, run a task, and understand approval prompts.
Learn how the dsh launcher separates profile flags from app arguments and how headless mode runs one persisted task without a server.
Trace how profiles, bundles, cordis.patch.yml files, and command overlays compose the plugin tree that dsh boots.
Create a minimal TypeScript plugin, insert it with a patch file, and understand cleanup and dependency injection from the official tutorial.
Follow one task through the agent loop and learn when session, agent, and capability events are the right extension point.
Understand the three runtime surfaces that turn a model response into a durable, tool-using agent task.