DeepSeek-TUI: A DeepSeek Coding Agent in the Terminal
DeepSeek-TUI, with 2.3k GitHub stars, is a terminal coding tool written in Rust. Optimized for DeepSeek V4, it supports a 1M-token context, parallel sub-agents, and three operation modes.
DeepSeek-TUI: A DeepSeek Coding Agent in the Terminal
DeepSeek-TUI, with 2.3k GitHub stars, is a terminal coding tool written in Rust. Optimized for DeepSeek V4, it supports a 1M-token context, parallel sub-agents, and three operation modes.
DeepSeek-TUI is a terminal coding agent written in Rust -- you can think of it as a "DeepSeek edition of Claude Code". It runs in the terminal, is specifically optimized and adapted for DeepSeek V4, and has already picked up 2.3k stars on GitHub.
DeepSeek-TUI running in the terminal
Core Features
Chain of thought visible in real time: DeepSeek-TUI streams the model's reasoning process straight into the terminal -- how the model analyzes the problem, which path it takes, whether it changes its mind along the way, all visible as it happens.
1M-token context: The project uses DeepSeek V4's full 1 million token context window by default. When the context is close to full it compresses automatically, and the compression strategy is designed around DeepSeek's prefix caching mechanism.
RLM multi-sub-agent mode: One main model directs up to 16 V4 Flash subtasks running at the same time, useful for batch analysis or task decomposition. Flash output costs roughly one-third of Pro, so handing subtasks that do not need heavy reasoning to Flash cuts the overall bill considerably.
RLM multi-sub-agent parallel mode
Three operation modes:
| Mode | Description | Best for |
|---|---|---|
| Plan | Read-only exploration; produces a plan without executing | Understanding a codebase, making plans |
| Agent | Every tool call requires your confirmation (default) | Day-to-day development, step-by-step review |
| YOLO | Fully automatic, no interruptions | Batch tasks, high-trust operations |
Hands-On Experience
Strengths
- Almost everything Claude Code can do, it can do too: reading and writing files, executing shell commands, searching the web, managing Git, orchestrating sub-agents, connecting to MCP servers -- full feature coverage
- A caching strategy tuned for DeepSeek: when compressing the context it tries to preserve the stable earlier portion so the cache keeps hitting
- Skills installation support: an extension mechanism similar to Claude Code's
- Friendly for users in China: a Chinese README, TUNA Cargo mirror support, and release packages that can be hosted on Alibaba Cloud OSS or Tencent Cloud COS
Caveats
- Cache hit-rate issues: with many sub-agents running, cache hit rates are hard to guarantee. Tokens that miss the cache cost 10x the price of ones that hit
- Per-turn cost display: worth watching on long sessions so the bill does not exceed expectations
Installation and Usage
One-line install:
npm install -g deepseek-tuiPrecompiled binaries are available for Linux, macOS, and Windows.
API configuration: supports DeepSeek's official API, NVIDIA NIM, Fireworks, self-hosted SGLang, and several other routes.
Project URL: github.com/Hmbown/DeepSeek-TUI
Use Cases
- Developers doing day-to-day work with the DeepSeek API
- Scenarios involving extremely long code files (1M-token context)
- Batch code analysis and parallel multi-task processing
- A DeepSeek-based alternative to Claude Code
Toolin Editorial Team
Categories
Related articles
JoyAI-Echo: An Open-Source Framework for 5-Minute Long-Video Generation
JD's open-source AI long-video framework generates cross-shot audio-video up to 5 minutes in a single pass, supports local edits, and says goodbye to gacha-style rerolling.
OpenSquilla: A Token-Saving Middleware Layer for Agents
An open-source Agent Harness framework that cuts Agent token costs by more than 50% through intelligent routing, context management, and self-evolution.
Gemma 4 12B: Run a Multimodal AI Model on a 16GB Laptop
Google releases a 12-billion-parameter open-source multimodal model with text, image, and audio input; it runs locally on a laptop with just 9GB of VRAM, under the Apache 2.0 license.
Hermes Desktop: The Open-Source Agent Moves onto the Desktop
Nous Research launches Hermes Desktop, an open-source desktop agent covering macOS/Windows/Linux that reuses the CLI agent's full skills and memory — usable with just mouse clicks.
Kimi Work: A Local AI Agent for Knowledge Workers
Moonshot AI launches a desktop general-purpose Agent with Agent clusters, browser control, and financial data sources, letting office workers handle their daily work with AI.
OpenClaw 2026.6.1: Native Windows Support at Last
The world's largest open-source AI Agent project ships a major update: native Windows support, a Skill Workshop that lets Agents self-evolve, and multi-Agent Workboard collaboration — 1.6 billion PCs become compute nodes.