Documentation Index
Fetch the complete documentation index at: https://agents.candu.ai/llms.txt
Use this file to discover all available pages before exploring further.
Layer 1: Configuration
Your config lives athttps://cdn.candu.ai/agentikConfig/{clientToken}/. Three files are deployed together: actions.json (action registry), agent.json (system prompt customisation), and wizard configs like segment-wizard.json.
Config changes propagate via CDN — typically within 60 seconds of save. Versioning lets you roll forward or back without redeploying your app.
Layer 2: Candu Actions API
Handles the conversation layer: receives user input, calls the AI provider, translates responses into tool calls. Fully managed. When the AI decides to call an action, it returns a structured tool call to the SDK.Layer 3: Clientside SDK
Runs inside your application, inside your users’ browsers. The First-party Snippet is the critical piece. It intercepts tool call events from the SDK and executes the actual HTTP request using yourfetchFn and your credentials.
What Candu touches and what it doesn’t
| Candu | Your app | |
|---|---|---|
| Conversation state | Managed by Candu | — |
| AI model calls | Managed by Candu | — |
| API execution | Never | Your snippet, your credentials |
| Auth tokens | Never sees them | Injected by your fetchFn |
| Identifiers | Not persisted | Injected into AI prompt at request time, then discarded |
| API response bodies | Recorded in execution logs (as JSON) | Your API determines what’s returned |
| Execution logs | actionId, inputs, outputs, status | — |