Introduction

Foldspace embeds an AI agent in your web app through a single SDK. The agent reads a user’s request in natural language, then acts on it: it calls actions you define, navigates your UI, fills and submits forms, and renders interactive components inline in the conversation.

You configure what the agent can do in Agent Studio — actions, input schemas, knowledge, and navigation — then wire those actions to your application code with the SDK. The agent decides when to call an action; your handler runs it. It can also read live page context (shared state) so its responses match what the user is doing right now.

How it fits together

• SDK: a script you drop into your frontend (SPA or MPA). It loads asynchronously and renders the agent.
• Agent Studio: where you define actions, input schemas, knowledge, and behavior. Your Agent Key lives here.
• Actions: typed capabilities you implement in your app. The agent decides when to call them; your code executes them.
• Chatterblocks: interactive UI (forms, cards, dashboards) the agent renders inside the conversation.
• Observability: every session, action call, latency, and token cost is tracked.

Where to go next

• Setup: install the SDK and show the agent.
• User context: identify users and enrich analytics.
• Implementing actions: wire actions to your app logic.