Deepseekeyes

Key features

• Image understandingpaste an image → the configured multimodal model reads the original pixels → DeepSeek receives validated evidence and answers in the same task.
• discover the desktop, then observe only the target window to reduce irrelevant pixels
• Connect local stdio servers or remote Streamable HTTP endpoints.
• User images pass through ctx.attachments.readImage() as the original Harness ImageBlock.
• Browser controlask DeepSeek to open a page → Browser Computer Use observes, opens, scrolls and clicks → every action returns a fresh state so DeepSeek can verify the result or recover from a missing target.