Now with Claude Code support

Canvas UI for
Claude Code

Visualize, orchestrate, and understand AI agents on an infinite canvas. Watch Claude Code work in real-time across multiple sessions.

Demo video coming soon

Install in seconds

Run the server locally to connect your Claude Code sessions.

Terminal
$ npx @agent-canvas/server start

Or install globally:

$ npm install -g @agent-canvas/server
$ agent-canvas start

No config needed

Works out of the box

Local only

Data stays on your machine

Quick start

Running in under a minute

Turn agent chaos into clarity

Stop juggling terminal windows. See all your AI agents working together in one unified view.

Infinite Canvas

Arrange multiple agent sessions on a limitless workspace. Zoom, pan, and organize your AI workflow exactly how you want.

Real-time Visibility

Watch agents think and work in real-time. See tool calls, file edits, and terminal commands as they happen.

Claude Code Integration

Native support for Claude Code CLI sessions. Connect existing terminals or spawn new agents directly from the canvas.

Session Management

Track multiple conversations, branch explorations, and maintain context across your entire development workflow.

Get started in seconds

No complex setup. No configuration files. Just launch and go.

01

Launch the Canvas

Open Claude Canvas in your browser. The infinite canvas awaits your agents.

02

Connect Your Agents

Spawn new Claude Code sessions or connect existing terminals. Each agent appears as a node on your canvas.

03

Watch & Orchestrate

See your agents work in real-time. Resize, reposition, and interact with each session from one unified view.

Built for real workflows

Whether you're shipping features or exploring new codebases, Claude Canvas adapts to how you work.

Multi-repo Development

Work on frontend, backend, and infrastructure simultaneously. Each agent handles a different part of your stack.

Agent 1: React components | Agent 2: API endpoints | Agent 3: Database migrations

Code Review & Refactoring

Have one agent analyze code while another implements improvements. Compare approaches side by side.

Agent 1: Security audit | Agent 2: Performance optimization

Learning & Exploration

Explore unfamiliar codebases with multiple agents tackling different areas. Build understanding faster.

Agent 1: Architecture overview | Agent 2: Testing patterns | Agent 3: Dependencies

Ready to see your agents clearly?

Claude Canvas runs entirely in your browser. No accounts, no tracking, no data collection. Your agent sessions stay on your machine.