The creative canvas for AI agents

Brainy Canvas gives agents a visible workspace for creative work: create projects, add typed nodes, connect models, run workflows, and leave a canvas that humans can inspect. It is built for agents first, then creators, students, builders, and marketers who want the same visual power without coding.

Start free Explore templates

Agent quick start

export BRAINY_CANVAS_API_KEY=bc_live_...

curl -H "Authorization: Bearer $BRAINY_CANVAS_API_KEY" \
  https://brainycanvas.com/api/cli/node-types

Use the same API key from Claude Code, Codex, OpenClaw, Hermes Agent, custom scripts, or any agent runtime that can call HTTP endpoints.

Why agents need a canvas

Prompt logs are not enough when agents produce creative work. A canvas turns invisible chains into inspectable artifacts: inputs, model choices, branches, generated files, and final outputs.

142+

AI models available for workflows

36+

providers across text, media, and vision

API

keys and CLI routes for agent access

Use Brainy Canvas with your agent stack

The agent surface is intentionally simple: create an API key, give it to the agent, and let the agent build a workflow that remains visible to the user.

Claude Code

Use Brainy Canvas from Claude Code through MCP where configured, or through shell/API calls with a Brainy Canvas API key.

claude mcp add brainy-canvas https://your-brainy-canvas-mcp.example.com/mcp

Codex

Give Codex an API key and let it create projects, add nodes, connect workflows, and run model branches from the terminal.

export BRAINY_CANVAS_API_KEY=bc_live_...

OpenClaw

Use Brainy Canvas as the visible creative workspace behind agent-generated content, research, and transformation tasks.

export BRAINY_CANVAS_API_KEY=bc_live_...

Hermes Agent

Any agent with HTTP tools can call the CLI API to build canvases, upload inputs, run workflows, and fetch outputs.

curl -H "Authorization: Bearer $BRAINY_CANVAS_API_KEY" https://brainycanvas.com/api/cli/node-types

The agent workflow pattern

Agents can use Brainy Canvas as a shared memory and review layer for multimodal work.

1. Create a project for the task.
2. Add source, planning, and output nodes.
3. Connect nodes into a visible workflow.
4. Run one node or the full workflow.
5. Inspect outputs on the canvas before publishing or handing off.

Project state stores nodes, edges, viewport, and outputs.

Node descriptors expose the available capabilities for dynamic tools.

Run endpoints support node-level and workflow-level execution.

Human reviewers can open the canvas and understand what the agent did.