AI Agent Hub
Back to plugins
deepseek-harness-split-screen preview

deepseek-harness-split-screen

Client Updated 2026.08.25

Run the following command in DeepSeek Harness:

dsh plugin install syncended/deepseek-harness-split-screen

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install syncended/deepseek-harness-split-screen in your terminal; the full source code is available at https://github.com/syncended/deepseek-harness-split-screen

About this plugin

The DeepSeek Harness Web GUI is a single-session surface by default, so juggling multiple workspaces, prompt comparisons, or parallel projects means constant tab-hopping and sidebar re-selection. deepseek-harness-split-screen brings an iTerm-style multiplexing workflow into the Harness browser view: lay out up to twelve independent panes side-by-side or stacked, each pointing at a different workspace and session without interfering with the others.

Every pane carries the full live experience out of the box — streaming transcript, session status, queued replies, a stop button, and history paging — plus a compact model and access-mode selector and a clickable context ring that surfaces used/remaining/limit, turn and step counts, token throughput (tok/s), input/output totals, and cache-hit metrics. Drag a divider to resize, drag one pane header onto another to swap entire tab stacks, and every layout ratio, pane order, and unsent draft persists in browser localStorage so a refresh restores your workspace exactly.

Built for developers who run several Harness workspaces concurrently on the same machine, need to compare outputs from different models or context windows side-by-side in the browser, or want a one-click jump back to the main view for structured approvals and plan reviews. It slots into a registered center-workspace occupant and leaves the protocol layer untouched; uninstalling it restores the native conversation surface immediately.

Screenshots

Use Cases

  • Run several workspace sessions in parallel and compare outputs side-by-side on one screen
  • Compare responses from different models or context windows without constant tab-hopping
  • Track token usage, streaming output, and queued replies in each pane in real time

Best For

  • Developers who run multiple Harness workspaces concurrently on the same machine
  • Engineers who need to A/B prompt or model outputs side-by-side in the browser
  • AI users who want a terminal-style multi-pane workflow inside the Web GUI