A frontier coding agent. Writes. Debugs. Fixes. Refactors. All from your terminal — and gets sharper with every session. Just prompt. It handles the rest.
A frontier coding agent. Writes. Debugs. Fixes. Refactors. All from your terminal — and gets sharper with every session. Just prompt. It handles the rest.
A CLI built for speed. And taste.
//A CLI built for speed. And taste.Beautiful. Fast. Crafted to the highest standards. Every interaction feels instant. Every output, crisp.
One-Shot Setup
//One-Shot Setup.One command and you are in. Install globally and start coding in seconds. Auto-updates and smooth onboarding flow with no configuration required.
~/project
Self-Updating
//Self-Updating.One command — `cmd update` — fetches the latest release and swaps it in. No lockfile churn, no broken globals, no re-auth.
~/project
Beautiful UI and Clean UX
//Beautiful UI and Clean UX.Terminal UI with crisp typography, thoughtful colors, and clean layouts. Designed with care, it delivers a focused developer experience that looks as good as it works.
~/project
Blazing Fast
//Blazing Fast.Sub-second startup, instant responses, and zero lag. Command Code is engineered for speed so you stay in flow.
~/project
Tiny Footprint
//Tiny Footprint.The lightest frontier coding agent. A single binary, ~40 MB resident memory, no Electron, no heavyweight IDE. Runs comfortably on laptops, CI, and remote shells.
~/project
Multi-Directory Workspace
//Multi-Directory Workspace.Point at a sibling repo, a shared package, or an infra folder with `--add-dir` or `/add-dir`. The agent indexes them as first-class workspace roots — ideal for monorepos and polyrepos.
~/project
Your taste. Learned. `npx taste`
//Your taste. Learned. `npx taste`.Every accept, reject, and edit becomes a signal. Your patterns. Your micro-decisions. Learned continuously.
Continuous Taste Learning
//Continuous Taste Learning.Learns from every accept, reject, and edit. Your preferences, naming, patterns, and style are learned automatically and written back as project-level skills in `.commandcode/taste/`.
~/project
Push & Pull Taste
//Push & Pull Taste.Share your coding taste with teammates, pull taste from the community. Namespace-based packages with public/private visibility.
~/project
Taste Onboarding from Other Agents
//Taste Onboarding from Other Agents.Already using Claude Code, Cursor, or another agent? Command Code learns your existing preferences with a single command.
~/project
Taste from Any Repo
//Taste from Any Repo.Point Command Code at any local path or GitHub repo and it distills the coding taste — patterns, naming, structure — into a reusable package you can publish or apply.
~/project
A mode for every moment
//A mode for every moment.Interactive chat. Headless automation. Plan-only analysis. Auto-accept for speed. Background sandbox for the heavy stuff.
Interactive Mode
//Interactive Mode.Real-time chat interface with your AI coding agent. Ask questions, edit files, run commands, all in a conversational flow.
~/project
Headless Mode
//Headless Mode.Non-interactive single-query mode for CI/CD pipelines, scripts, and automation. Supports piped stdin.
~/project
Plan Mode
//Plan Mode.Analyze and plan without executing any changes. Read files, search code, discuss architecture. No modifications.
~/project
Auto-Accept Mode
//Auto-Accept Mode.Skip confirmation prompts for rapid implementation. Toggle instantly with shift+tab.
~/project
Background Sandbox
//Background Sandbox.Fire-and-forget. `cmd sandbox "…"` spins up an isolated task that runs to completion in the background and reports back. Resume any session with `--id`.
~/project
Pro tools. Built in.
//Pro tools. Built in.File ops, shell, grep/glob, extended thinking, todo tracking, web search, `/review`. The full stack of dev tools — first-class, all the way.
File Operations
//File Operations.Read, write, edit files and directories. Glob pattern search and regex grep across your entire codebase.
~/project
Shell Execution
//Shell Execution.Execute shell commands with real-time output streaming. Run tests, builds, linters, and any CLI tool directly.
~/project
Extended Thinking
//Extended Thinking.Deep reasoning space for complex architectural decisions, refactoring strategies, and multi-step problem solving.
~/project
Agentic /review
//Agentic /review.Deep, multi-dimensional PR review — correctness, conventions, test coverage, and clarity, scored and explained. Pulls PR context automatically, extracts existing comments, and posts back structured feedback.
~/project
Web Search & Fetch
//Web Search & Fetch.Search the web and fetch pages directly from the agent. Up-to-date docs, API references, and release notes, grounded into your session.
~/project
Todo Tracking
//Todo Tracking.Long tasks get broken into a live todo list the agent maintains itself — every step visible, every completion marked. Great for multi-file refactors and migrations.
~/project
Hackable. Out of the box.
//Hackable. Out of the box.Connect external tools, databases, APIs through the Model Context Protocol. Add `/skills`, `/commands`, `/agents`. Built to be remixed.
MCP Server Integration
//MCP Server Integration.Connect to external tools and data sources via HTTP or stdio transport. Tool discovery, OAuth authentication, and dynamic scoping.
~/project
Agent Skills
//Agent Skills.Lightweight, self-contained instruction sets for specialized tasks. User-level and project-level skills with progressive disclosure.
~/project
Custom Slash Commands
//Custom Slash Commands.Define reusable commands with markdown templates. User-level and project-level, with variable substitution.
~/project
Sub-Agents
//Sub-Agents.Spawn focused sub-agents for research, refactors, and parallel exploration. Keeps the main context clean, runs work in isolated scopes.
~/project
It remembers. So you don't have to.
//It remembers. So you don't have to.Project instructions, coding guidelines, architecture notes. Remembered. Compacted. Applied — every session.
AGENTS.md Project Memory
//AGENTS.md Project Memory.Define project-level instructions, code style guidelines, and architecture notes. Automatically loaded every session.
~/project
Checkpoints & Rewind
//Checkpoints & Rewind.Automatic snapshots before every modification. Rewind to any point. Restore code, conversation, or both.
~/project
Session Resume & Compact
//Session Resume & Compact.Continue where you left with --resume. Long conversations auto-compact to keep context fresh and focused.
~/project
Any frontier model. Any provider. Zero lock-in.
//Any frontier model. Any provider. Zero lock-in.Anthropic. OpenAI. Open-source. Sonnet, Opus, GPT-5, GLM, Kimi, DeepSeek, MiniMax. Switch on the fly. No strings attached.
Model & Provider Selection
//Model & Provider Selection.Switch between AI models on the fly — Sonnet 4.6, Opus 4.7, Haiku 4.5, GPT-5.4, and more. Pick the right model for the task.
~/project
Open-Source Models
//Open-Source Models.Run GLM-5, Kimi K2.5, and MiniMax M2.5 alongside Sonnet, Opus, and GPT. Bring your own provider, zero lock-in, full control over every request.
~/project
Even better, together.
//Even better, together.`/share` conversations. `/pr-comments` for review context. `npx taste push/pull` to sync taste across the team. All without leaving the terminal.
Conversation Sharing
//Conversation Sharing.Generate shareable links for any conversation. Share context, decisions, and code changes with your team.
~/project
Organizations & Teams
//Organizations & Teams.Create organizations, manage team members, and share taste packages. Org-level usage tracking and billing.
~/project
PR Context & Comments
//PR Context & Comments.`/pr-comments` pulls every review comment on the current branch into the session — CI failures, line-level feedback, requested changes — grounded into the agent before it touches a line.
~/project
Until coding feels like thinking.
//Until coding feels like thinking.Keyboard shortcuts. IDE integration. Fine-grained permissions. Live diagnostics. Designed to keep you in flow.
Keyboard Shortcuts
//Keyboard Shortcuts.Esc to rewind, shift+tab to toggle modes, alt+p for model selection, ctrl+t to toggle learning feed, ctrl+o for expanded tool output.
~/project
IDE Integration
//IDE Integration.VSCode extension for context-aware coding. Current file, selection, cursor position, all sent to the agent automatically.
~/project
Permission & Trust System
//Permission & Trust System.Fine-grained permission controls. Standard mode for per-file confirmation, trust mode for known projects, plan mode for read-only analysis.
~/project
Status & Diagnostics
//Status & Diagnostics.`/status`, `cmd info`, and the diagnostics tool give the agent — and you — a full view of auth, model, git, PR, workspace, and skills state at a glance.
~/project
Studio — your Command Code control plane
//Studio — your Command Code control plane.Manage usage, billing, taste packages, personal and organization settings from the Command Code Studio.
Studio Dashboard
//Studio Dashboard.Real-time usage tracking, token consumption breakdown, cost analysis, and historical data, all in one view.
API Key Management
//API Key Management.Authenticate via browser OAuth, or generate, rotate, and scope API keys from Studio. Org-level keys for team access.
~/project
Usage & Credits
//Usage & Credits.Per-session credit accounting, premium-request tracking, and monthly breakdowns. Check usage without leaving the terminal.
~/project
Taste Library
//Taste Library.Browse, star, and manage your taste packages in Studio. Public for the community, private for your team. Every package is versioned, diffable, and one command away.
~/project
//Take Command of your code.Ship 10x faster with the same team, less time, and your coding taste. Install, sign in, and start coding.