Prompts library
The prompts module stores and serves reusable AI prompt templates with versioning and access control.
What you'll learn
- Why a centralized prompt store
- Authoring + versioning prompts
- Consuming prompts from your app or agent
Why a centralized prompt store
Inline prompts everywhere = drift, no review, no rollback. Centralizing gives you:
- Versioning — every change is a new version; you can pin or roll back
- Variables — prompts are templates with typed placeholders
- Permissions — RBAC on who can read / edit
- Observability — see which prompt got called, when, by whom
Authoring
Web UI: Prompts → New Prompt.
A prompt template:
Title: customer-support-triage
Variables: { customerMessage: string, customerHistory: string[] }
Body:
You are a triage agent. Read the customer message below and history.
Output JSON: { category, urgency, suggestedReply }.
Customer message:
{{customerMessage}}
History:
{{#each customerHistory}}- {{this}}\n{{/each}}
Consuming a prompt
import { BurdenoffClient } from '@burdenoff/sdk-node';
const client = new BurdenoffClient({ token });
const rendered = await client.prompts.render('customer-support-triage', {
customerMessage: 'My order hasn't arrived',
customerHistory: ['Placed order on May 10', 'Asked status on May 15'],
});
// Now pass `rendered` to your LLM provider
Versioning
client.prompts.render('customer-support-triage@latest', vars); // default
Next steps
- AI plugin development — exposing prompts as tools