Plugins

Vision for text-only models: paste an image and the model switches to a Vision Toolkit variant for image Q&A, multi-image comparison, long-screenshot OCR, screenshot-to-UI reproduction, element grounding, and pixel diff. No API key by default — images are processed by the author-hosted free service, 100 per machine per day; configurable to your own provider.

Let DeepSeek Harness see and operate any program on Windows. The model looks at the real screen and drives the mouse, keyboard and windows itself. It is fast: several steps run in a single round trip, and a window can be focused before input lands, so the click hits the right one. It can do anything you would do on a desktop, including but not limited to CAD drawing, circuit simulation, and sending messages.

Accessibility-first macOS computer use: fresh observations, stale-state rejection, scoped permissions, and safe input.

AX-first macOS computer use for DeepSeek Harness: accessibility snapshots with stable refs, background input, Unicode typing, menus and apps, set-of-mark screenshots, on-device OCR and verified actions — 11 tools bridged to the macos-cu CLI.

Windows Computer Use for DeepSeek Harness with 21 tools and verified targeting: controls are addressed through their UI Automation identity and patterns, and numbered marks are checked against the live window before input, so a stale target is refused instead of clicked. A known sequence runs in one computer_batch call, one model round trip; incremental observations, compact receipts (normally no more than about 6 KiB extra), bounded handoff waits (HANDOFF_WAIT_TIMEOUT/HANDOFF_HELD_TIMEOUT as continuations), the desktop-block cycle boundary and a deterministic 100 s owner-scoped leave backstop keep context small and prevent a stuck handoff from waiting out the tool deadline. Approval modes settle an unanswerable ask honestly: in a danger-full-access session with no approval card the action runs under that grant, every other sandbox — and every delegated subagent session — refuses and names the unavailable channel instead of reporting a user rejection, and gate diagnostics land in policy.jsonl. Human handoff, a model-visible stop brake and prebuilt native workers.

AX-first macOS computer use for DeepSeek Harness: accessibility snapshots with stable refs, background input, Unicode typing, menus, apps, set-of-mark screenshots, on-device OCR, verified actions, and optional Jev (TypeSafe System One) semantic guards — 1

Hand a command-line-verified operation over to the GUI: drive it once, verify with ground truth, and hand back a self-contained illustrated tutorial.

Plug-in vision for text-only models on DSH, with native interaction for image understanding and generation, GUI automation, through layered evidence memory and cache.

Lightweight local visual grounding engine exposed as agent tools

Computer Use enhancement plugin for DeepSeek Harness (DSH). 纯 JS 视觉差分、网格位置编码、长程任务闭环与人机共驾授权。