Back to directory

dsh-tui

Curated pickMaintenance: Active

ccch1mneyyy/dsh-tui

Claude Code 风格全屏终端 TUI:像素鲸鱼顶栏、双流光大字、实时工作状态行、思考流式展开、双击 Esc 时间回溯、上下文进度条 + TPS 仪表;接入 Agent preset(standard/code/minimal/cordis)、Skills、MCP、Goals、Todos、子代理,/ 命令注册表与随附技能(audit/bug/review/practice/pr_comments/release-notes/vuln-check);纯插件挂载零核心改动。

View on GitHubHomepage
$ npm install -g @deepseek-ai/dsh @deepseek-harness-tui/dsh-tui

2,832

stars

158

forks

TypeScript

Language

MIT

License

2026-08-13

Created

2026-09-04

Last push

MIT-licensed Claude Code-style terminal TUI for DeepSeek Harness, mounted as a pure dsh profile plugin (`dsh --profile dsh-tui`) with zero core changes — whale top bar, live agent status, context progress, TPS gauge and double-Esc time rewind.

DSH integration

Compatible

Author-claimed

Safety audit

Unaudited

Last verified

2026-08-21

License

MIT

01What can it help you accomplish?

  • Run DeepSeek Harness coding-agent sessions in a Claude Code-style fullscreen terminal TUI

    A polished TUI experience: pixel whale splash bar, dual flowing-glow title, real-time working status line, streaming thought expansion, double-Esc time rewind, blue-white context progress bar + TPS gauge

    Developers who run the official dsh CLI in a terminal and want a beautiful, observable Claude Code-style TUI instead of plain output

  • Monitor agent state, token usage and context budget in real time during long DSH sessions

    Live working status, segmented context progress, TPS, cache hit rate, reasoning level, input/output tokens and Git/session info in the status line, with virtualized rendering that keeps long sessions cheap

    Power users and operators running long DeepSeek Harness sessions who need visibility into context budget and throughput

  • Manage full DSH session workflows — resume, rewind, fork, export — from the terminal

    Session resume/new/compact/export, `/btw` side questions, model switching, and session rewind/fork via double-Esc; all local commands route through official DSH links

    Everyday dsh users who want session resume, time rewind, fork and Markdown export workflows without leaving the terminal

02How to install into DeepSeek Harness

Prerequisites

  • An available terminal TTY, the official `dsh` CLI, and `pnpm` 10+ (first run auto-initializes the dsh-tui profile, which needs pnpm)
  • `DEEPSEEK_API_KEY` in order to run models

Installation steps

  1. 01

    Install the CLI and the plugin globally: `npm install -g @deepseek-ai/dsh @deepseek-harness-tui/dsh-tui`

    $ npm install -g @deepseek-ai/dsh @deepseek-harness-tui/dsh-tui

  2. 02

    Launch with `dsh-tui` — the first run auto-initializes the dsh-tui profile; alternatively run `sh install.sh` (repo root, wraps the profile install with a pnpm pre-check) or `dsh plugin --profile dsh-tui add @deepseek-harness-tui/dsh-tui`

    $ dsh plugin --profile dsh-tui add @deepseek-harness-tui/dsh-tui

Verify the integration

Not specified by the author

03DSH integration and capability boundaries

DSH integrationCompatible

Pure plugin mounted on a dsh profile — `dsh plugin --profile dsh-tui add @deepseek-harness-tui/dsh-tui`, after which `dsh-tui` is equivalent to `dsh --profile dsh-tui`; zero core changes

  • Terminal-native interaction

    Keyboard, mouse (optional fullscreen mode) and clipboard in a terminal TTYStreaming Markdown, structured tool cards, `/` command and `@` file completion, image attachments (PNG/JPEG/WebP/GIF as persistent blocks), history search, message selection, inline/alternate-screen rendering and `/lang` CN-EN UI switching

    Ctrl+V paste reads the system clipboard via platform tools (PowerShell Get-Clipboard on Windows, osascript/pbpaste on macOS, wl-paste/xclip/xsel on Linux); clipboard bitmaps are written into the attachment store
  • Observable agent state

    Live session/event stream from the DSH servicesReal-time working status, segmented context progress bar, TPS gauge, cache hit rate, reasoning level, input/output tokens and Git/session info

  • Session workflow: resume / rewind / fork / export

    Existing DSH sessions and the current conversation`/resume` session browser, `/new`, `/compact`, `/export` Markdown, `/btw` side questions, model switching, and double-Esc session rewind/fork

    `/model` switching forks the session (DSH has no in-place model switch); the choice is persisted to `~/.dsh-tui/model.json`, and the default preset to `~/.dsh-tui/agent-preset.json`
  • DSH official capability wiring

    A dsh profile with the DSH services and command registryAgent presets (standard/code/minimal/cordis + liangshen), Skills, MCP servers, Goals, Todos, sub-agents and `ask_user_question` questionnaires, all connected through existing DSH services or the registry

    MCP servers are mounted via `@deepseek-ai/dsh-mcp-client` and register tools as `mcp__<server>__<tool>`After startup the TUI checks npm in the background for new versions; `/update` updates the runtime in the dsh-tui profile and restarts the session

04Who is it for? When not to use it?

Good for

  • Developers who run the official dsh CLI in a terminal and want a beautiful, observable Claude Code-style TUI instead of plain output
  • Power users and operators running long DeepSeek Harness sessions who need visibility into context budget and throughput
  • Everyday dsh users who want session resume, time rewind, fork and Markdown export workflows without leaving the terminal

Not for

  • dsh-TUI implements no sandbox of its own — it uses the current dsh profile's file, Shell, sandbox and approval policies; on Windows there is no sandbox backend, so the composition falls back to `danger-full-access` without approval prompts. Check the profile config before launching in environments with sensitive credentials or untrusted repos.

05Compatibility, maintenance and safety notes

  • The project is marked public beta (status badge in the README); behavior and interfaces may change between releases.
  • dsh-TUI implements no sandbox of its own — it uses the current dsh profile's file, Shell, sandbox and approval policies; on Windows there is no sandbox backend, so the composition falls back to `danger-full-access` without approval prompts. Check the profile config before launching in environments with sensitive credentials or untrusted repos.
  • `Ctrl+V` clipboard paste depends on external tools per platform (PowerShell Get-Clipboard on Windows, osascript/pbpaste on macOS, wl-paste/xclip/xsel on Linux); paste degrades or is unavailable when the tools are missing or the session is unreachable.
2026-08-132026-08-19v0.8.6

MIT · actively maintained (latest release v0.8.6, 2026-08-20)

06Frequently asked questions

How does dsh-TUI connect to DeepSeek Harness — native plugin or MCP?

It's a pure plugin mounted on a dsh profile: `dsh plugin --profile dsh-tui add @deepseek-harness-tui/dsh-tui` (or the wrapped `sh install.sh`), after which `dsh-tui` is equivalent to `dsh --profile dsh-tui`. Agent presets, Skills, MCP, Goals, Todos and sub-agents are wired through existing DSH services and the command registry — the TUI is the frontend, not an MCP wrapper.

What prerequisites do I need?

An available terminal TTY, the official `dsh` CLI, and `pnpm` 10+ (the first run auto-initializes the dsh-tui profile and needs pnpm). Running models also requires `DEEPSEEK_API_KEY`.

Does it modify the dsh core?

No — zero core changes, pure plugin mounting. Installing the plugin enables it; uninstalling leaves no core patches behind.

Where do model calls and session data go?

The TUI only handles interaction and rendering. The session log is the source of truth; model calls, tool execution, fork/resume, compaction and persistence stay owned by the DSH services. After startup the TUI also checks npm in the background for new versions, and `/update` applies them to the profile runtime.

What are the main limitations?

It's public beta. dsh-TUI has no sandbox of its own — it uses the current dsh profile's file/shell/sandbox/approval policies, and on Windows the composition falls back to `danger-full-access` without approval. Also note: `/thinking` is not persisted, `/compact` is unavailable under the minimal preset, and model switching works by forking the session.

08Data and sources

  • Author-claimedgithub.com5f0ddf7673c4…

    零核心改动,纯插件挂载。安装插件即可启用,卸载后不会留下核心补丁。

  • Author-claimedgithub.com5f0ddf7673c4…

    sh install.sh # 或:dsh plugin --profile dsh-tui add @deepseek-harness-tui/dsh-tui # 之后 dsh-tui 与 dsh --profile dsh-tui 等价

  • Author-claimedgithub.com5f0ddf7673c4…

    - **DSH 官方能力接入**:Agent preset、Skills、MCP、Goals、Todos、子代理、 `ask_user_question` 问卷都通过现有服务或注册表连接。

This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-21. Found an error? Submit a correction.

🏆

Best DeepSeek Harness Plugins

Twelve plugins worth installing first — picked from the whole catalog, across every category.

DSH Plugins is an independent community directory of DeepSeek Harness plugins. Not affiliated with or endorsed by DeepSeek. Third-party plugins are not security-audited — review the source before installing.

New DeepSeek Harness plugins, weekly. No spam.