OpenClaw Canvas & A2UI

The Canvas System is an agent-driven visual workspace in OpenClaw. Agents can render structured UI components declaratively rather than returning plain text.

A2UI Framework

A2UI (Agent-to-UI) is the declarative framework that enables agents to emit UI component descriptions as part of their responses. The Canvas interprets these declarations and renders them in the control interfaces (Web UI, macOS app).

Key properties:

• Agent-driven: agents decide what to render, not just what to say
• Declarative: agents describe UI structure; the Canvas handles rendering
• Integrated: part of the response delivery phase, not a separate system

Use Cases (inferred)

• Displaying structured data (tables, charts, code blocks with metadata)
• Interactive elements agents surface to the user
• Multi-step workflow visualizations

Related

• OpenClaw Gateway — response delivery phase where Canvas output is returned
• Agent Loop (OpenClaw) — the loop that generates A2UI output
• OpenClaw — entity overview