Shipped • MIT Licensed

The agent workspace
framework.

A shipped canvas where agents and humans share the same structured panels under governed rules.

What it is

Agentable is the shipped agent workspace framework. It combines a tldraw-based canvas with a governed panel system so AI agents and people work on the same surface.

Canvas

Bounded or infinite whiteboard substrate powered by tldraw.

Panels

Four tiers of validated, composable UI components.

Agents

Agents open, fill, compose and mutate panels safely.

Why agents need a shared surface

Chat-only interfaces force agents to work with unstructured JSON blobs. Agentable gives agents a structured, validated UI surface where they can act with precision while humans remain in control through explicit approval gates.

This shared workspace turns agent actions into visible, reviewable work instead of hidden function calls.

Core capabilities

Production-ready primitives for agent-native interfaces.

Panel registry

IR + validator + catalog for all panel types

DataAdapter boundary

Clean separation between agent logic and UI state

Validation pipeline

Schema-driven validation for every panel mutation

HITL mutation gates

Human-in-the-loop approval for all agent actions

World model digest

Multi-agent shared understanding of canvas state

Voice + chat

Native multimodal input for agents and users

Package exports

Ready-to-use entry points for different integration patterns.

Package	Description	Best for
whiteboard	Full canvas with panel system	Primary workspace hosts
react-canvas	React component wrapper	Deep React integrations
embed	Lit web component	Career concierge & third-party embeds
career-pack	Domain-specific panel library	Career and recruiting workflows

Structured UI, not JSON blobs

Unlike chat-only UIs, Agentable gives agents a real interface to work with.

Chat-only interfaces

• Unstructured JSON responses
• No visual context for agents
• Difficult human review

Agentable

• Structured, validated panels
• Visual shared workspace
• Explicit HITL approval flows

Architecture overview

View full architecture documentation →

See the AI-native canvas

Drop a single Lit web component into any page to mount the Agentable whiteboard substrate — an AI-native canvas design framework where agents and people share panels on an infinite canvas.

Embed snippet

<section class="agentable-demo">
  <agentable-canvas
    tenant="product-demo"
    primary-color="#2563EB"
    canvas-mode="infinite"
    welcome-message="Hello from Agentable"
  ></agentable-canvas>
</section>
<script type="module" src="https://moss-career-blueprint.pages.dev/embed/agentable-canvas.js"></script>

The real tldraw whiteboard host

Beyond the chat-and-panel embed above, Agentable also ships an infinite tldraw canvas host — the same production whiteboard substrate agents and people share panels on. Try the live host below, or drop it into your own page with an iframe.