Home + Sidebar
Projects, sessions, and useful information, in their place.
1.0.0 RC2 is here ↗THE TERMINAL, REIMAGINED.
Desktop-class feel.
Terminal-sized.
Your AI coding workbench in the terminal. Rich interactions. Immediate feedback. A smoother path from planning to code review.
macOS / Linux / WindowsBUILT WITH DART + NOCTERM
01 / FEEL AT HOME
Every interaction. Worth getting right.
Real buttons, mouse interaction, and a workbench you can arrange. A focused terminal with room for complex work.
Plan + Review
Read plans alongside the conversation. Inspect your changes.
Automatic TLDR
Keep the full answer. Get the essentials when you return.
Coding Plan Usage
See supported provider quotas and usage in your workbench.
BUILT-IN GIT CLIENT
Built-in Git client. From changes to commits, in one place.
Browse changes in the file tree and compare them side by side. Stage a file or a hunk, check what belongs in the commit, and generate a commit message. Review and commit your work directly in Crux.
Side-by-side diffStage files or hunksGenerate commit messages
Explore the workbench and Git review ↗MERMAID, IN YOUR TERMINAL
A terminal can show more than text.
Ask the agent to explain architecture, flows, and state changes with Mermaid. Read diagrams alongside the conversation; drag to explore an overflowing diagram.
Diagram guide ↗
Agent-built plugins. Live in your workbench.
Tell the agent what you need. It can generate small, interactive plugins on the fly and display them in the Sidebar, Home, or both. New plugins and changes load automatically, without restarting Crux.
YOU
“Add a gold-price plugin to my workbench.”
Read status, click buttons, and run quick actions right in the plugin. Ask the agent to change its content or interactions as your needs evolve.
Create your first plugin ↗Interact directly. Keep the task moving.
Choose a city, refresh the weather, or complete and submit a form. The agent builds an interactive interface; your actions feed back into the conversation and move the task forward.
Choose a cityRead the forecastRefresh the weather
Explore GenUI ↗03 / LESS WAITING
Time and context. Spent where they matter.
Index in seconds. Help the agent find key code.
Semble indexes code in seconds and helps the agent find key implementations by meaning, reducing repeated searches and irrelevant context.
Reuse cached input. Lower token costs.
Project usage has reached 99% cache hit. Stable prefixes reuse repeated input to lower token costs; savings depend on the model, task, and provider’s cache pricing.
More work in one round.
Independent tool calls run in parallel, reducing model round trips and token usage from repeated context.
Open your terminal. Start your next thing.
DeepSeek · Kimi · MiniMax · Codex · Custom compatible APIs
Install and set up ↗Read the manual ↗