DeepSeekEyes: Auditable Vision and Computer Use Plugin for DeepSeek Harness
DeepSeekEyes is a DeepSeek Harness (DSH) plugin developed by dttxorg under the MIT license, designed to address the issues that image pixels cannot be directly used as model input in agent tasks and the lack of evidence traceability. This plugin constructs an auditable visual and cross-platform Computer Use runtime, with core features including: supporting multimodal models to read raw pixels to provide unscaled, traceable visual evidence; offering native desktop control and browser automation operations on Windows/macOS; and integrating an MCP application layer to restrict, hash, and audit tool invocation results. Additionally, it provides health checks, failover, strict evidence validation, and token usage visualization capabilities. By default, automation and MCP features are disabled and require manual configuration to enable.
Read Moredsh-computer-use: Adding Virtual Mouse for Human-Like Operations to DeepSeek Harness
dsh-computer-use is a DSH workflow plugin maintained by 988hj7tczd-oss, providing 11 model-friendly tools such as screen_observe and computer_click for harness-desktop based on cua-driver. It first observes the screen using the AX tree, then performs pixel-level operations like clicking, typing, and scrolling through an independent virtual cursor across macOS/Windows/Linux. This article introduces the core capabilities, install.sh installation steps, security configurations, and typical usage.
Read MoreAdd macOS Computer Control to DeepSeek Harness with dsh-computer-use
DSH-Computer-Use is a DeepSeek Harness tool plugin maintained by Anionex, which provides native macOS computer control for intelligent agents. It conducts real-time observations based on the accessibility tree, rejects expired states, divides read and write permissions by application, and tries its best not to move the system cursor or inject pointer events into the global desktop. This article checks the installation commands, Skill loading methods and capability boundaries against the community directory page and GitHub README. Currently, it only supports macOS 14 and above, with the early version 0.1.0.
Read More