vCode v1.5 is here

vCode just got its biggest upgrade yet.

Your local machine. Cloud intelligence. Now safer, smarter, and fully extensible. vCode connects Lloyd to your own computer — the AI thinks in the cloud, but every action runs locally on your machine, in your repos, with your tools.

Read the vCode guide
curl -fsSL https://vcode.artificiallabs.ai/install | sh
vCode terminal interface with streaming actions and local file edits
Safety first

You're always in control

Approve, audit, and rewind anything Lloyd does on your machine.

Granular permissions

vCode asks before it edits files or runs commands. Set allow/deny rules once and share them with your team by committing them to the repo.

Permission modes

Choose cautious default, focused acceptEdits, read-only plan, or --yes when you just want it done.

Checkpoints & /rewind

Every file the agent touches is snapshotted first. One command rolls anything back. Experiment fearlessly.

Effortless terminal

A terminal experience that feels modern

The comforts of a modern editor inside a native terminal workflow.

Modern line editor

Command history, tab completion for commands and file paths, and multiline input — all built in.

@-mention anything

Type @src/server.js to attach a file, or @screenshot.png to share an image. Lloyd can literally see your UI.

Pick up where you left off

vcode --continue reopens your latest session instantly. Sessions live in the cloud, so nothing is lost when your laptop sleeps.

/doctor

One command checks your connection, auth, and version so you never debug your tools instead of your code.

Knows your project

An agent that learns your codebase

Teach vCode once, benefit forever — across every project and session.

Memory hierarchy

Personal notes in ~/.vcode/VCODE.md, team conventions in VCODE.md, and private scratch notes in VCODE.local.md. Start a note with # mid-session and save it on the spot.

/init

New to a codebase? vCode explores it and writes its own briefing document so you hit the ground running.

Custom slash commands

Drop a markdown file in .vcode/commands/ and it becomes a reusable /command for your whole team.

/model switching

Switch between available AI models mid-session whenever the task demands it.

Plan mode

/plan puts Lloyd in read-only mode: it analyzes and proposes, you approve, then it builds.

Built to extend

Extensible, the industry-standard way

Plug into the tools and workflows you already use.

MCP support

Connect any Model Context Protocol server with vcode mcp add. Its tools appear to Lloyd automatically and run securely on your machine — GitHub, databases, internal APIs, anything in the MCP ecosystem.

Hooks

Run your own shell commands before or after every agent action. Enforce lint gates, auto-format on save, trigger notifications — a failing pre-hook blocks the action entirely.

Scriptable

Pipe input in, get JSON or streaming JSON out. vCode slots straight into your CI and automation.

Your local shell

Your local shell and filesystem

vCode executes commands in your own terminal emulator, with your local prompt and project directory. It reads and edits files on your actual machine, not a remote sandbox.

  • Runs in the terminal you already use
  • Sees your exact local environment, tools, and secrets
  • Works with your dotfiles, aliases, and shell history
vCode executing a shell command in the user's local terminal
vCode install command and OS badges for Windows, macOS, and Linux
One binary

One binary, every OS

Install vCode with a single curl command or Homebrew tap. The same binary runs on Windows, macOS, and Linux.

  • curl -fsSL https://vcode.artificiallabs.ai/install | sh
  • brew install vcode
  • Windows, macOS, and Linux builds available
Steer from anywhere

Steer from anywhere

Send context, corrections, and approvals from Slack, the web panel, or your phone. vCode keeps working on your machine while you stay in control.

  • Web panel with live session stream
  • Slack notifications and steering messages
  • Mobile-friendly status and approvals
vCode checkpoint timeline showing rewind and restore points
vCode web panel and phone steering a live local session
Checkpoints

Checkpoints and rewind

vCode automatically saves checkpoints as it works. If an edit goes wrong, rewind to a known-good state with a single command.

  • Automatic checkpoints before risky edits
  • /rewind to any previous point
  • Restore files, working tree, and session state
Get started in seconds

Two commands, then you're building

vcode login
vcode

Download vCode for Windows and Linux from your panel's vCode page. Already installed? vCode tells you when an update is available at startup.

vCode by AiLabs — the coding agent that works where you work.

What's new

Release highlights

Four releases of shipping the features that make vCode a professional-grade coding agent.

v1.5.0 — Extensible

  • MCP servers: vcode mcp add/list/remove connects local Model Context Protocol servers; use /mcp to see connected servers and tools.
  • Hooks: define PreToolUse, PostToolUse, and Stop shell hooks in .vcode/settings.json.
  • Image attachments: @image.png uploads images so vision-capable models can see your UI.

v1.4.0 — Programmable

  • Memory hierarchy: user, project, and local memory files with @import support.
  • # note quick-add, /memory editor, and /init project bootstrap.
  • Custom slash commands from .vcode/commands/*.md with $ARGUMENTS.
  • /model switching and /plan read-only mode.

v1.3.0 — Comfortable

  • Modern line editor with history, tab completion, and multiline input.
  • @file inlining, --continue, /doctor, and /terminals.
  • Update notifications and --output-format stream-json for automation.

v1.2.0 — Safe

  • Permission engine with allow/deny rules, four permission modes, and interactive approval prompts.
  • File checkpoints with /rewind restore.
  • Shareable project settings in .vcode/settings.json.

Stop reviewing code that might work.
Start shipping code that does.

Get Lloyd on your local machine today.