Your AI Workspace,
infinite
canvas.
Agent Grid is the workspace where you manage Claude Code, Codex, terminals, a code editor, source control, browsers, and notes as panes on an infinite canvas. One place to orchestrate everything.
Infinite canvas
& workspaces.
Pan, zoom, and arrange panes freely. Snap to grid, auto-organize with one click, and switch between project workspaces in tabs — each with its own folder and layout.
- Multi-tab workspaces, one per project
- Drag-and-drop a folder to set the workspace context
- Auto-organize button to fit panes neatly
- Minimap and pane counter for orientation
Claude Code
as an agentic pane.
Anthropic's coding agent runs as a fully agentic pane on your canvas — it reads your repo, edits files, runs commands, and uses MCP tools, all without leaving your workspace.
- Reads, edits, and writes files in your project
- Runs shell commands and dev scripts
- MCP tool integration out of the box
- Permission modes: Default, Plan, or Bypass for CI
Master Claude
delegates the work.
When a task is too big for one agent, a coordinator Claude spawns and orchestrates worker Claudes in parallel via MCP — fanning out the work and collating the results into one thread.
- Spawns and broadcasts to worker panes via MCP
- Workers run in parallel, each with their own context
- Aggregates worker results back into the master thread
- Visual edges connect master to its workers on the canvas
Codex
for a second opinion.
OpenAI's coding agent as a pane. Run it side-by-side with Claude on the same task to compare approaches, or pick the agent whose strengths fit the job.
- Run Codex and Claude on the same task and diff their work
- Switches in seamlessly via the spawn menu
- Same canvas, same workspace, different brain
- Auto-detects the codex CLI on your system
2. Create middleware/rateLimit.ts
3. Wire it into the login route
QA Agent
runs your tests.
A test-focused agent that plans, runs, and reports on your test suite — backed by Claude or Codex. Get a clean breakdown of what passed, what failed, and why.
- Plans test coverage from your spec files
- Executes the suite and captures screenshots on failure
- Reports failures with reproduction steps
- Pick Claude or Codex as the backend brain
Terminal
no transpiled fakery.
A real shell session backed by node-pty and rendered by xterm.js. Your bash, zsh, or fish — full ANSI colors, full keybindings, no surprises.
- Pseudo-TTY via node-pty (it's a real shell)
- 256-color ANSI rendering with xterm.js
- Inherits your env, profile, and shell of choice
- Pin one per workspace, or spawn many
Code Editor
for direct edits.
Edit project files in a VS Code pane right next to your agents and terminals. No alt-tab, no context loss — your editor lives on the canvas where the work happens.
- VS Code rendered as a pane
- File tree, syntax highlighting, search
- Open any file in your workspace
- Sits side-by-side with the agent that's editing it
Source Control
with AI commits.
A Git pane with stage, diff, branch switching, and history — plus a single button to ask Claude to write a clear, conventional commit message for what you're about to ship.
- Stage, unstage, and revert from the UI
- Visual diff with syntax highlighting
- Branch switching, history, and reflog
- AI-generated commit messages on demand
Browser
preview without leaving.
A Chromium webview pane. Preview localhost while your dev server runs, browse docs, or keep an auth session scoped per workspace — all on the canvas.
- Full Chromium webview in a pane
- Sessions are isolated per workspace
- Opens localhost, docs, or any URL
- DevTools available when you need them
Notes that
think with you.
Dictate with Whisper, auto-polished by Claude. Embed Mermaid diagrams inline. Agents read your notes and can work through them or tag them.
- Whisper voice dictation in any note
- Claude auto-polishes filler words and structure
- Mermaid diagrams render inline as you type
- Agents can read, reference, and tag your notes
Title
label your canvas.
When the canvas grows, you need landmarks. Title panes are auto-scaling section headers that group related agents and panes into named regions you can find at a glance.
- Auto-scaling text fills the pane
- Pick from a palette of subtle colors
- Drop above any group to label it
- Doesn't get in the way — purely visual
Ready to orchestrate?
Get early access to Agent Grid and manage all your AI agents from a single, beautiful, persistent workspace.
Be the first to know when Agent Grid launches.