DeepSeek Harness Plugins
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Adds an OpenCode Go model sync panel to DSH settings, fetching all live models, probing each for availability and vision support, and writing usable ones back to the config.
Auto-fills missing model fields (context window, max tokens, reasoning efforts, image input) from models.dev for custom provider models, applying user-defined defaults for unmatched entries without overwriting existing values.
CodeBuddy gateway inference plugin for dsh: dynamic multi-model sync, web search/fetch, image generation, OpenAI-compatible upstreams, and TraeWork credit channel.
Model-driven context rollover that replaces LLM summarization with deterministic checkpoints, preserving durable state across fresh windows.
A multi-agent job-matching workflow plugin that produces traceable, evidence-based company fit reports grounded in your confirmed preferences.
Integrates Qoder subscriptions into DeepSeek Harness with text and image input, streaming, reasoning, and tool calls across Global and China services.
Schedule one-off or Cron-based tasks for the DeepSeek Harness Web Profile, running on time or during off-peak idle windows.
Universal MCP bridge plugin for DeepSeek Harness that connects agents to any MCP server over stdio/SSE/Websocket with automatic tool, resource, and prompt registration.
Native Chrome control plane for DSH agents: reuse your existing profile and signed-in tabs with semantic observation, verified actions, virtual pointer visibility, and instant human takeover.
Hold-to-talk voice input for the DeepSeek Harness Web composer with local SenseVoice (sherpa-onnx) recognition — offline, no API key, audio never leaves your machine.
Discards inherited pending prompts on session fork so the child session responds to your new message instead of a leftover from the source.
A project-level MCP auto-loading plugin for DSH that mounts server configs on session open, supports hot-reload, and scopes tool visibility per session.