Terax Ai
Terax Ai - Lightweight (7MB) Terminal-first AI-native dev workspace
Last verified:
What is Terax Ai?
Terax Ai is a terminal-first AI-native development workspace that combines a native terminal, code editor, file explorer, source control, web preview, and agentic AI side-panel into a single ~7-8 MB desktop application. Built on Tauri 2 with a Rust backend and React 19 frontend, it offers a lightweight alternative to Electron-based IDEs with a 300 ms cold start and no telemetry.
Key features include a WebGL-rendered xterm.js terminal with multi-tab support and split panes, a CodeMirror 6 editor with real Vim mode and AI autocomplete, hunk-by-hunk AI edit diffs for reviewing agent changes before they touch disk, and an AI Composer with voice input, custom agents, plan mode, and snippets. The app supports Bring Your Own Key (BYOK) for cloud AI providers like OpenAI, Anthropic, Google, Groq, Cerebras, and xAI, or runs fully offline via LM Studio, MLX, or Ollama.
Terax is designed for developers who want a fast, keyboard-first terminal with AI built in rather than bolted on. It's particularly suited for developers working with local LLMs, those who value privacy (no accounts, no telemetry, API keys stored in OS keychain), and developers who prefer a minimal footprint without the Electron tax. The app is free and open-source under Apache-2.0 license.
Terax Ai pricing
Pricing model: Freemium
Free and open-source under Apache-2.0 license. No paid plans, no upsell, no account required. You bring your own API keys for cloud AI providers (OpenAI, Anthropic, Google, Groq, Cerebras, xAI, OpenRouter, DeepSeek, Mistral) which are billed by those providers separately, or use fully local/offline models via LM Studio, MLX, or Ollama at no additional cost. API keys are stored securely in the OS keychain, never on disk in plaintext.
Terax Ai pros
- Only 7-8 MB on disk with no Electron tax
- 300 ms cold start speed
- No accounts, no sign-up required
- No telemetry ever
- Open-source under Apache-2.0 license
- BYOK supports OpenAI, Anthropic, Google, Groq, Cerebras, xAI, OpenRouter, DeepSeek, Mistral
- Fully offline support via LM Studio, MLX, or Ollama
- Real Vim mode with motions, registers, and marks
- WebGL-rendered xterm.js terminal for smooth scrolling
- AI edit diffs with hunk-by-hunk review before changes touch disk
- Voice input to talk to agents while coding
- Auto-detects dev servers like Vite, Next, Astro for web preview
- Custom agents with their own system prompt and tool subset
- Plan mode for multi-step work with confirmation before execution
- TERAX.md per-project memory versioned with your repo
- 10 bundled editor themes including Nord, Tokyo Night, Catppuccin, Gruvbox
- Background image with adjustable opacity and blur
- Ripgrep-class project-wide search speed
- Native PTY backend via portable-pty
- Split panes for horizontal and vertical terminal layouts
Terax Ai cons
- Not a full IDE - lacks debugger UI and refactoring tools
- Not code-signed yet on Windows (shows 'Windows protected your PC' warning)
- Built mostly by one person (crynta), smaller team than major IDEs
- No hosted service - entirely local desktop app
- Wayland rendering glitches on Linux may require WEBKIT_DISABLE_DMABUF_RENDERER workaround
- AppImage on Linux needs FUSE installed
- AI provider policies apply to prompts (provider may retain data)
- Relatively new project with only 5.6k stars compared to established IDEs
Frequently asked questions about Terax Ai
What is Terax?
Terax is a lightweight terminal-first AI-native dev workspace that bundles a native terminal, code editor, file explorer, source control, web preview, and agentic AI side-panel into a single ~7-8 MB window. It runs on macOS, Linux, and Windows with no telemetry and no account required.
Is Terax free?
Yes, Terax is completely free and open-source under the Apache-2.0 license. There are no paid plans, no upsell, and no account required. You only pay for API keys if you use cloud AI providers, or you can run fully offline with local models at no cost.
What AI providers does Terax support?
Terax supports BYOK (Bring Your Own Key) for OpenAI, Anthropic, Google (Gemini), Groq, xAI (Grok), Cerebras, OpenRouter, DeepSeek, Mistral, and any OpenAI-compatible endpoint. For local/offline use, it supports LM Studio, MLX, and Ollama.
How do I install Terax?
Download the appropriate build for your OS from the website or GitHub Releases: macOS (Apple Silicon .dmg or Intel .dmg), Linux (AppImage, .deb, .rpm, or AUR via yay -S terax-bin), or Windows (x86_64 NSIS installer). Terax auto-updates from GitHub Releases.
Does Terax collect my data?
No. Terax has no accounts, no analytics, and no backend. API keys are stored in the OS keychain, never on disk in plaintext. Settings and session state are stored locally. Network requests only happen when you explicitly trigger them (AI calls, update checks, web preview).
Can I use Terax offline?
Yes, Terax can run fully offline through LM Studio, MLX, or Ollama with local LLMs. Your code stays on disk and nothing leaves your machine when using local models. You can also use Terax as a plain terminal without any AI functionality.
What is the TERAX.md file?
TERAX.md is per-project memory and configuration that gets versioned with your repository. It allows AI agents to maintain context about your project across sessions, storing project-specific memory and configs that agents can reference.
Does Terax have Vim mode?
Yes, Terax has first-class Vim mode with full support for motions, registers, and marks. You can bring your existing dotfiles. The editor is built on CodeMirror 6 with sub-millisecond keystroke latency.
What themes are available in Terax?
Terax includes 10 bundled editor themes (Atom One, Aura, Copilot, GitHub Dark/Light, Gruvbox Dark, Nord, Tokyo Night, Xcode Dark/Light) and 10 app palettes. You can also build custom themes in-app, export/share them as JSON, and use background images with opacity and blur sliders.
How do AI agents work in Terax?
Terax's agentic AI side-panel can read files, run commands, propose changes, browse files, and multi-edit with approval gating. Every edit lands in a reviewable diff before touching disk. Agents support sub-agents, custom agents with their own system prompts, snippets/skills for reusable workflows, plan mode for multi-step work, and voice input.